Luminous Landscape Forum
Site & Board Matters => About This Site => Topic started by: John Camp on August 12, 2011, 07:50:38 pm
-
Given the the controversy over the past couple of weeks, it would be interesting to see an expanded article that would summarize (and simplify) the various ETTR possibilities -- or perhaps two or three articles summarizing different views. As I have mentioned in my comments, I'm not so much interested in an engineering discussion, as in a "prescriptive" article or set of articles, that would deal with the when and how-to aspects of ETTR...if it's to be used at all. The various arguments have left me seriously uncertain about procedure...
JC
-
Given the the controversy over the past couple of weeks, it would be interesting to see an expanded article that would summarize (and simplify) the various ETTR possibilities -- or perhaps two or three articles summarizing different views. As I have mentioned in my comments, I'm not so much interested in an engineering discussion, as in a "prescriptive" article or set of articles, that would deal with the when and how-to aspects of ETTR...if it's to be used at all. The various arguments have left me seriously uncertain about procedure...
JC
Well, I'll kick off with a few methods that have worked for me.
(1) First one should attempt to get the histogram and 'highlight flashing' as close as possible to the ideal RAW histogram, bearing in mind that the histogram is based upon a jpeg from the RAW data.
To do this, in my experience, requires adjusting the camera's jpeg settings so that the review image looks very lack-lustre and unappealing, so much so that you wouldn't want to show people the image you'd just taken in case they thought you didn't know what you were doing, or in case they got the impression their P&S was sooo.. much better than your expensive DSLR.
Such adjustment of the camera's jpeg settings will need a bit of experimentation. You might find you need to set contrast, saturation and sharpening at a minimum so that the 'highlight warning' doesn't flash when the image is still underexposed, from the RAW perspective.
(2) Buy a new camera which has a linear noise and dynamic range response (on the graphs at DXO Mark) that reduces by one f stop (or EV) for every doubling of ISO, so that there is no fundamental image quality advantage to increasing ISO.
The Nikon D7000 and Pentax K5 are in this category. With such a camera, you have only one worry regarding ETTR, and that's inadvertent overexposure at base ISO.
Unfortunately, the review image with this method can look even worse than the first method, if you've been underexposing at base ISO instead of using a higher ISO. It may be so dark you can hardly see it at all and you may have to flatly refuse to show it to anyone in order to protect your reputation.
(3) Take the time and trouble to obtain a correct exposure for ETTR purposes before taking the shot. One effective way of doing this, using the camera's in-built exposure meter, is to set the camera to spot metering mode and, using a single focussing point, take a reading of the brightest part of the image you want to retain detail, then increase exposure by approximately 3 stops, depending on the model of your camera. Again you will have to experiment to determine whether that increase is 2.5 stops, 3 stops or 3.5 stops.
(4) Bracket either exposure or ISO. This is my favourite method because I can keep the review image looking bright and sparkly to impress the beautiful female subject I've just photographed.
There may be other useful techniques for achieving ETTR, but those are the ones I've tried and that work for me.
-
You should purchase the Camera to Print and Screen video series, it contains Michael's current view on this subject and more. There is a segment on ETTR with Jeff showing examples of noise at both 'normal' exposure and ETTR exposures. You can see the IQ differences, they're very material.
I'm really enjoying the video series so far - 5 installments are available to view with more coming online over the next several months. I feel that CPS is the most well thought out of all the Michael and Jeffrey series. The ETTR segment is no exception. I can see hours of preparation go into making each video segment. Care was taken to make clear lesson goals delivered in a conversational manner. Side by side comparisons, zooming in on image quality issues and basically explaining and showing at the same time are really great instructional tools. This series is exceeding my expectations - it's both a learning and viewing experience. I think you will find it enjoyable too. Excellent editing helps - hat tip to Chris.
-
Well, I'll kick off with a few methods that have worked for me.
(1) First one should attempt to get the histogram and 'highlight flashing' as close as possible to the ideal RAW histogram, bearing in mind that the histogram is based upon a jpeg from the RAW data.
To do this, in my experience, requires adjusting the camera's jpeg settings so that the review image looks very lack-lustre and unappealing, so much so that you wouldn't want to show people the image you'd just taken in case they thought you didn't know what you were doing, or in case they got the impression their P&S was sooo.. much better than your expensive DSLR.
Such adjustment of the camera's jpeg settings will need a bit of experimentation. You might find you need to set contrast, saturation and sharpening at a minimum so that the 'highlight warning' doesn't flash when the image is still underexposed, from the RAW perspective.
(2) Buy a new camera which has a linear noise and dynamic range response (on the graphs at DXO Mark) that reduces by one f stop (or EV) for every doubling of ISO, so that there is no fundamental image quality advantage to increasing ISO.
The Nikon D7000 and Pentax K5 are in this category. With such a camera, you have only one worry regarding ETTR, and that's inadvertent overexposure at base ISO.
Unfortunately, the review image with this method can look even worse than the first method, if you've been underexposing at base ISO instead of using a higher ISO. It may be so dark you can hardly see it at all and you may have to flatly refuse to show it to anyone in order to protect your reputation.
(3) Take the time and trouble to obtain a correct exposure for ETTR purposes before taking the shot. One effective way of doing this, using the camera's in-built exposure meter, is to set the camera to spot metering mode and, using a single focussing point, take a reading of the brightest part of the image you want to retain detail, then increase exposure by approximately 3 stops, depending on the model of your camera. Again you will have to experiment to determine whether that increase is 2.5 stops, 3 stops or 3.5 stops.
(4) Bracket either exposure or ISO. This is my favourite method because I can keep the review image looking bright and sparkly to impress the beautiful female subject I've just photographed.
There may be other useful techniques for achieving ETTR, but those are the ones I've tried and that work for me.
I would add this:
(5) Change your RAW converter, and/or expose appropriately for which ever RAW converter you use.
-
I would add this:
(5) Change your RAW converter, and/or expose appropriately for which ever RAW converter you use.
If anyone knows of a RAW converter that can recover more highlight detail and color than ACR (to a worthwhile degree, that is), I'd like to hear about it.
My favourite converter used to be RSP (Raw Shooter Premium). I simply preferred the rendition of microcontrast and color, and the arrangement of the controls. However, I did notice that recovery of highlights was not quite as good as ACR. I also found it easier to get more natural skin tones in ACR.
Each converter tends to have its strengths and weaknesses, and the fact one is simply familiar with the controls and use of one particular converter is itself a strength, from one's own perspective.
In two of the methods of achieving ETTR I've outlined above, (adjusting the camer's jpeg settings for contrast and color saturation, and/or taking a spot meter reading of the brightest area in the scene then increasing exposure by a set amount), it is necessary to experiment first using one's converter of choice to find out what settings to the jpeg image, and/or what degree of increase in exposure provide the most accurate result with regard to assessing the correct exposure for ETTR.
-
If anyone knows of a RAW converter that can recover more highlight detail and color than ACR (to a worthwhile degree, that is), I'd like to hear about it.
My favourite converter used to be RSP (Raw Shooter Premium). I simply preferred the rendition of microcontrast and color, and the arrangement of the controls. However, I did notice that recovery of highlights was not quite as good as ACR. I also found it easier to get more natural skin tones in ACR.
Each converter tends to have its strengths and weaknesses, and the fact one is simply familiar with the controls and use of one particular converter is itself a strength, from one's own perspective.
In two of the methods of achieving ETTR I've outlined above, (adjusting the camer's jpeg settings for contrast and color saturation, and/or taking a spot meter reading of the brightest area in the scene then increasing exposure by a set amount), it is necessary to experiment first using one's converter of choice to find out what settings to the jpeg image, and/or what degree of increase in exposure provide the most accurate result with regard to assessing the correct exposure for ETTR.
That's exactly my point. ACR is among the best, if not the best, at highlight recovery, and, if that is the program of choice, ETTR makes more sense. With RPP, blown highlights aren't easily fixed, but bringing up exposure in the converter is quite a bit better, so ETTR isn't really necessary, especially if the trade off is color issues.
One should expose with their converter of choice in mind.
-
In two of the methods of achieving ETTR I've outlined above, (adjusting the camer's jpeg settings for contrast and color saturation, and/or taking a spot meter reading of the brightest area in the scene then increasing exposure by a set amount), it is necessary to experiment first using one's converter of choice to find out what settings to the jpeg image, and/or what degree of increase in exposure provide the most accurate result with regard to assessing the correct exposure for ETTR.
One should expose with their converter of choice in mind.
mmm if we agree that proper ETTR means no highlight clipping at all, I don't see why our particular RAW developer should have anything to say in the process of achieving ETTR in the camera. Proper ETTR is the same for all RAW converters: maximum RAW exposure before clipping, and all of them should produce similar results with perfectly ETTR'ed captures since no highlight recovery strategies are involved. I think this is again the discusion about using a RAW converter as something it was never intended to be, a RAW analyzer.
Regarding the settings for the most accurate histogram in the camera, modifying contrast, saturation, sharpness,... in the camera are peccata minuta (in particular a contrast curve should never produce/prevent any clipping). Modifying them will not help much since the most differentiating process between the RAW histogram and camera's histogram is by far white balance. So the way to make them look closer should be cancelling WB (http://www.luminous-landscape.com/forum/index.php?topic=22250.0) first.
Regards
-
Regarding the settings for the most accurate histogram in the camera, modifying contrast, saturation, sharpness,... in the camera are peccata minuta (in particular a contrast curve should never produce/prevent any clipping). Modifying them will not help much since the most differentiating process between the RAW histogram and camera's histogram is by far white balance. So the way to make them look closer should be cancelling WB (http://www.luminous-landscape.com/forum/index.php?topic=22250.0) first.
Regards
Guillermo,
In my experience, changing the in-camera jpeg settings for constrast and saturation does have a significant effect on the appearance of the histogram.
Again, the histogram of different models of cameras may behave differently to the adjustments of the in-camera jpeg settings, but my experience with the Canon 5D was that the default jpeg settings, when shooting a typical landscape scene for example, might result in the entire sky flashing with the 'clipped highlight' warning, on the camera's LCD screen, despite the fact that such apparently clipped sky could be brought back to normal in ACR with the appropriate adjustments, and with no excessive and unnatural degree of cyan, too.
However, with contrast and saturation of the camera's jpeg set to a minimum, the histogram of a shot of the same landscape scene with same exposure would show just small areas of flashing in the brightest parts of the sky.
With such settings in place, one could then make a fairly accurrate assessment if the exposure had clipped some of the highlights. Small areas of 'flashing' indicated a good ETTR. No flashing at all indicated at least a small degree of underexposure. Moderate to large areas of flashing sky indicated that some clipping had taken place.
I'm speaking in the past tense because I no longer use this method. I prefer the less 'anal' approach of bracketing exposure or ISO.
I understand your UNIWB method would allow a slightly greater exposure to be used before clipping takes place. But such a method also results in a review image on the camera's LCD screen appearing yukky. It also might create additional problems for the user in post processing, correcting for such a WB which is way out by a significant degree.
Just how much trouble does one want to take for marginal improvements in shadow noise? If I have the luxury of time on my side (and a tripod), and the scene is contrasty, I'd rather bracket exposure and merge to HDR. If some movement has taken place during the bracketing, CS5 can sometimes handle it.
If I don't have a tripod but I'm really careful holding the camera steady, CS5 can usually handle the merging to HDR and correct for any minor misalignment.
Also, with cameras such as the D7000 and K5, ETTR is a non-issue above base ISO.
Regards
-
I understand your UNIWB method would allow a slightly greater exposure to be used before clipping takes place. But such a method also results in a review image on the camera's LCD screen appearing yukky. It also might create additional problems for the user in post processing, correcting for such a WB which is way out by a significant degree.
Ray, I think you are inventing "additional problems" where none exist.
-
Ray, I think you are inventing "additional problems" where none exist.
You mean it is not true that the review image will be yukky?
You mean that all converters and all users will have no problems perfectly correcting for the wrong WB?
-
You mean it is not true that the review image will be yukky?
You mean that all converters and all users will have no problems perfectly correcting for the wrong WB?
If you can't properly set a WB in your RAW editor of choice ... you probably shouldn't be shooting RAW ... or trying to ETTR ...
-
If you can't properly set a WB in your RAW editor of choice ... you probably shouldn't be shooting RAW ... or trying to ETTR ...
Never heard that one before. "Why are you shooting Jpeg?" "Oh! Because I can't properly set the WB in the RAW converter. It's always off, for some reason."
-
What's to revisit?
If manufacturers did their job and gave us cameras that could automatically expose raw so that the image is just below important clipping (with a user selectable tolerance), and then "normalize" the image on the rear LCD and in the raw file metadata, we would have the absolutely optimum exposure possible for each and every shot. No need for anything else.
Isn't that what I wrote in the article?
Michael
-
I'm speaking in the past tense because I no longer use this method. I prefer the less 'anal' approach of bracketing exposure or ISO.
+1
Thinking about what I’m actually doing in practice, it is more a kind of "ETTR light" approach. Means to seek for a max exposure barely before we see those flashing highlight warning & relevant clipping with the camera luminance histogram (in-camera jpeg settings basically left at standard).
Then, if there is time and I’m in the mood, to bracket exposure in the + EV direction. +EV because at this starting point it seems to me unlikely that there’s already true Raw clipping involved.
To be honest, as much as I like theoretical part, my feeling is that any strict(er) ETTR concept & claim is stuck between a rock and a hard place, i.e. improving sensor performance & dynamic range, and improving "zero noise" HDR blending & tone mapping techniques.
Peter
--
-
You mean it is not true that the review image will be yukky?
It will have a green cast, but so what? At least the histogram will be more accurate. Is there some reason you think you need a color-accurate preview on the LCD? It's not like the LCD's are highly calibrated, anyways. If you're judging color based on the LCD preview you're asking for trouble.
You mean that all converters and all users will have no problems perfectly correcting for the wrong WB?
Any RAW converter worth using will allow you to set the WB without regard to the in-camera WB setting, so I'm not really sure what you're getting at here. And I for one don't see the point in constantly adjusting in-camera WB when it can be adjusted far more easily at the computer. Even if I didn't use UniWB, I'd just leave the in-camera setting on something like daylight, so I'm always going to be setting WB in the RAW converter anyways. For the most part I'm going to choose WB based on the aesthetic look I want for the image anyways (and if I want a 'neutral' WB I'll just shoot a gray card). So yeah, I would agree with Jeremy, anybody who think in-camera WB is important for anything other than judging exposure doesn't fully understand RAW.
-
In my experience, changing the in-camera jpeg settings for constrast and saturation does have a significant effect on the appearance of the histogram.
And in my experience too, I never said it didn't. What I mean is that since WB is the main cause of camera vs RAW histogram disagreement, it doesn't make much sense fiddling with the saturation or contrast settings if we are using any non-neutral WB.
-
+1
BR
Erik
What's to revisit?
If manufacturers did their job and gave us cameras that could automatically expose raw so that the image is just below important clipping (with a user selectable tolerance), and then "normalize" the image on the rear LCD and in the raw file metadata, we would have the absolutely optimum exposure possible for each and every shot. No need for anything else.
Isn't that what I wrote in the article?
Michael
-
You mean that all converters and all users will have no problems perfectly correcting for the wrong WB?
Considering the white balance has zero effect on the raw data, maybe (maybe because there are some users who will always have problems <g>).
-
mmm if we agree that proper ETTR means no highlight clipping at all, I don't see why our particular RAW developer should have anything to say in the process of achieving ETTR in the camera. Proper ETTR is the same for all RAW converters: maximum RAW exposure before clipping, and all of them should produce similar results with perfectly ETTR'ed captures since no highlight recovery strategies are involved. I think this is again the discusion about using a RAW converter as something it was never intended to be, a RAW analyzer...
Regards
It's because some converters deal with boosting exposure better than others, so sacrificing color for ETTR isn't always necessary or appropriate if you're using RPP with most newer cameras, and the makers of that program will tell you the same. In fact, to quote Andrey Tverdokhleb, the designer of RPP, in regards to the A900 and RPP:
"...I meant using camera light meter (or external one if you care), but don't push histogram to the right and use camera light meter as it's intended, i.e. expose most important part of a picture around camera midpoint. ETTR had some reasons for old cameras with low DR - noise was too close to the midpoint and we had to do this to minimize it. With late cameras which have over 9 stops of DR ETTR is very harmful for colors - midpoint is the most colorful place in A900 gamut and noise is not an issue there any more. In A900 gamut slowly narrowing down from midpoint to shadows and very quickly narrowing down from midpoint to highlights. This means that brightest stop of the camera range has most of colors gone forever and they cannot be restored with negative exposure compensation. I'd say all color critical parts should be below top 1.5 stops. Veiling glare from lens and sensor are the culprits here...Squeezing scene with high DR into sensor or film range is a totally valid approach when needed, same as exposing for shadows. ETTR however assumes that it's always better to shift histogram to the right, even when your scene is only 6 stops wide and sensor is 9.5 stops wide. You probably already noticed before that slightly underexposed shots can be amazingly colorful even after exposure correction and I definitely noticed that ETTR shots can be very dull after correction even if there was no clipping. So my point is that ETTR is not always better and shouldn't be used unconditionally. The whole approach that you need to pay attention only to highlights is very limiting - what's really important is were we place critical part of a picture on a sensor range. This critical part can be anywhere - in shadows, highlights or in the middle and we should understand that moving it closer to the middle gray will improve it's appearance and try our best...Gray point is the base point which light meter is calibrated to. There used to be an old film 18% gray standard (2.5 stops from saturation point at the top), than at some moment all vendors switched to 12.7% (about 3 stops). They all round up them differently, so there is some small variations, f.e. it can be 12.5%. Then they switched to even lower values. My A900 is about 10% gray (3.33 stops), but it looks like there is slight variations in different camera samples. This all is based on green channel only (not sure if they use same kind of green as on sensor though).
...Now recent crop of cameras like A900, D3X, 5D2 went even farther. A900 calibrated to 10% gray and picture exposed around this number will definitely look dark. Most likely vendors just optimized light meters to place gray point in the best spot of a sensor, which is a good thing - they can afford this because sensors are better and shadows are much cleaner now. Bad thing here is that pictures need to be compensated now. In RPP this means Compressed exposure correction about 0.7-0.8 for A900. It absolutely must be corrected! Other converters do this behind a curtain and apply some exposure corrections automatically, usually badly and based on their own (mis)understanding, tone curve is getting involved so that's were all this confusion comes from. I completely disagree with this approach and in RPP you'll actually see picture exactly as it was captured and you need to define your correction specifically and make it default."
-
Douglas, that endless speech has already been explained 1000 times in the forums. To believe all that I just need a pair of RAW files where it can be proved that colours get altered after proper ETTR. I have a feeling most people talking about colour shifts because of ETTR are just using clipped RAW files without knowing, or incorrect processing (this could include RAW converters handling eposure inapropiately, like Tverdokhleb suggests).
Some years ago I coded a program called Zero Noise that builds an output image from several input images, taking the most exposed non-clipped pixels. I consider valid a pixel as long as all three channel values in it are below a given safety threshold with respect to RAW saturation point. Do you know what threshold I decided to use by default? 90% of saturation in linear scale, that is only 0.15EV below RAW saturation. And I am pretty sure I could use 95% and the result would be the same.
Regards
-
It will have a green cast, but so what? At least the histogram will be more accurate. Is there some reason you think you need a color-accurate preview on the LCD? It's not like the LCD's are highly calibrated, anyways. If you're judging color based on the LCD preview you're asking for trouble.
So what? No problem at all for me. But some of us interact with others when we go out shooting. I find people often ask to see some of the shots I've taken, especially if the shots are of them.
Any RAW converter worth using will allow you to set the WB without regard to the in-camera WB setting, so I'm not really sure what you're getting at here.
If there's no problem, then fine. I didn't write there was, just that there might be, and the reason I thought there might be is because in the thread on this subject, that Guillermo linked to above, there was mention that earlier versions of Photoshop did not allow for sufficiently extensive adjustments to be made to correct the WB of an image where a magenta filter had been used to equalise the levels of R, G & B.
I personally I'm not interested in spending more time than I need to when processing images. I usually get satisfactory results with camera set to Auto WB. If there's a quick and easy method of restoring the correct WB of a UNIWB image, then fine. Tell me. I''m all ears.
-
My general guideline is to know how the indicators on the camera relate to raw channel clipping. Unfortunately this relationship is not clear (this is where the camera vendors could help us). For my personal 5D Mark II, for example, the spot meter is placed about 3 1/3 stops below the raw green channel clipping, under common illumination conditions. So to ETTR, I point the spot meter at the brightest area in the scene I wish to photograph, and open up 3 stops more than the suggested exposure value. That's it. Then in postprocessing, I adjust the (software) exposure to whatever is needed to achieve the overall desired image brightness.
Technical users (comfortable with dcraw) can do controlled tests to check how their in-camera meter relates to raw channels clipping.
Non-technical users can also do controlled tests using a fixed scene (a target with tones, such as a ColorChecker or stepwedge can help). Take a set of bracketed exposures, and note the in-camera spot meter reading for each. Gradually increase the exposure. Afterwards, try to postprocess all of them to have the same overall image brightness. At some point, your brighter exposures will start losing highlight detail (because they clipped) or start having color shifts. That means you've gone too far. The brightest exposure that avoids these problems is the +delta you want to use.
Eric
-
For my personal 5D Mark II, for example, the spot meter is placed about 3 1/3 stops below the raw green channel clipping, under common illumination conditions.
Now thats sobering.....
-
For my personal 5D Mark II, for example, the spot meter is placed about 3 1/3 stops below the raw green channel clipping, under common illumination conditions.
(...)
Technical users (comfortable with dcraw) can do controlled tests to check how their in-camera meter relates to raw channels clipping.
For my classic Canon 5D I obtained the same headroom as you:
(http://www.guillermoluijk.com/article/spot/histomax5d.gif)
Lighting conditions were not Daylight or any normalized lighting but my sitting room's Tungsten though.
Regards
-
There's another implication of the UNIWB method which seems unhelpful in many situations.
As I understand it, but please correct me if I'm wrong, the UNIWB method allows for a greater exposure, without clipping highlights, than one would usually be able to use.
It's basically like taking a DSLR that has a true base ISO of 100, and giving it a true base ISO of 50. One consequently gets approximately (maybe) a one stop DR and SNR advantage.
Now this is fine if a slow shutter speed is okay. In a sense, it's a method of turning a 35mm DSLR into an MFDB with regard to DR and noise at base ISO.
The problems arise when slow shutter speeds are not appropriate, and one needs to either increase ISO or underexpose.
In these situations, the UNIWB, it seems to me, is of no value.
If you are not sure about this, let's consider a specific example. I need (or want) a shutter speed of 1/200th at F8. The lighting conditions determine that I need to go to ISO 400 to get an ETTR, using Auto WB.
Guillermo's method suggests I can use 1/100th sec exposure at such an aperture and ISO, and consequently achieve lower noise. But 100th sec is too slow. I need to use 1/200th sec.
So, Guillermo's method suggests that I can use 1/200th at ISO 800, without blowing highlights.
Now my question is, why should 1/200th at ISO 800, with UNIWB, be better than 1/200th at ISO 400, using the camera's Auto WB?
I hope no-one is getting a headache reading this.
Or to put it another way, why should 1/200th at ISO 400, with Auto WB and correct ETTR, be worse than 1/200th with UNIWB at ISO 400, underexposed by one stop?
-
Ray using UniWB in no way dictates your exposure, in fact you don't have to be using ETTR to benefit from it. UniWB gives you a more accurate indication of the actual RAW exposure for each color channel - nothing more, nothing less. If you don't see the value in that, maybe UniWB isn't for you. But your arguments "against" UniWB in this thread seem more than a little silly to me.
If you're happy with with Auto-WB then by all means keep using it. I would never leave such an important creative decision up to a camera's automatic processing, but that's just me.
-
As I understand it, but please correct me if I'm wrong, the UNIWB method allows for a greater exposure, without clipping highlights, than one would usually be able to use
(...)
Now my question is, why should 1/200th at ISO 800, with UNIWB, be better than 1/200th at ISO 400, using the camera's Auto WB?
Ray, I think you have some misconception about UniWB. UniWB doesn't allow greater exposure, it simply provides more reliable histograms. Shooting at a given ISO, aperture and shutter, will end with the same RAW data no matter if you use UniWB or any other arbitrary WB setting.
Regarding your example: 1/200 at ISO800 (with any WB), will be better than 1/200 at ISO400 (again, with any WB) because the second shot will have less visible noise (not much less though). Using or not using UniWB will simply help you to be aware with more precision if highlights are clipping.
On the other hand using some exposure or another will end in different motion blurring or DOF, and it's the photographer who has to choose exposure paremeters according to the entire application requirements. UniWB is just a tool to more accurately know how much you are exposing, not a guide who decides exposure.
Regards
-
If you are not sure about this, let's consider a specific example. I need (or want) a shutter speed of 1/200th at F8. The lighting conditions determine that I need to go to ISO 400 to get an ETTR, using Auto WB.
Fine, but that 'determination' has no ETTR basis.
Guillermo's method suggests I can use 1/100th sec exposure at such an aperture and ISO, and consequently achieve lower noise. But 100th sec is too slow. I need to use 1/200th sec.
Thus, you won't ETTR but underexpose by 1 stop (relative to ETTR).
So, Guillermo's method suggests that I can use 1/200th at ISO 800, without blowing highlights.
Yes, but as a second best choice! Best is to ETTR (nothing beats real photons), if you cannot do that then it's better to boost the ISO to 800 than just underexpose at ISO 400.
Do note that with these small 1 EV differences, the improvement can be subtle if any (depending on e.g. camera).
Now my question is, why should 1/200th at ISO 800, with UNIWB, be better than 1/200th at ISO 400, using the camera's Auto WB?
It's not the UNIWB but the resulting decision to safely allow more photons (which requires freedom to increase/maximize photon count) into the equation that makes a difference. The WB is just metadata, and of no real consequence for the Raw conversion.
Cheers,
Bart
-
Ray, I think you have some misconception about UniWB. UniWB doesn't allow greater exposure, it simply provides more reliable histograms. Shooting at a given ISO, aperture and shutter, will end with the same RAW data no matter if you use UniWB or any other arbitrary WB setting.
Regarding your example: 1/200 at ISO800 (with any WB), will be better than 1/200 at ISO400 (again, with any WB) because the second shot will have less visible noise (not much less though). Using or not using UniWB will simply help you to be aware with more precision if highlights are clipping.
On the other hand using some exposure or another will end in different motion blurring or DOF, and it's the photographer who has to choose exposure paremeters according to the entire application requirements. UniWB is just a tool to more accurately know how much you are exposing, not a guide who decides exposure.
Regards
Really! So it's all a storm in a tea cup? I had some sort of concept that due to the application of an automatic white balance, the green channel might clip whilst the red and blue channels were well underexposed.
I thought, by applying a UNIWB, the channel most susceptible to clipping would be brought back into line with the other channels, allowing for a greater exposure.
I had some concept that the blue or red channel might be around one stop underexposed when the green channel was at an ETTR, for example. Or in some cases, as in a sunset, the red channel would be at an ETTR when the other channels would be underexposed.
So I've got it all wrong, then. ;D
-
Ray,
What might be confusing you is that there are a few photographers who will use external color correction filters to equalize the channel histograms and in this case your concerns are valid. What everyone here is talking about for UniWB is just setting up a WB in the camera to make the in camera histograms better reflect the RAW histograms. As noted, this doesn't change the RAW data at all. There is one notable exception, I believe it is the D200 or some other Nikon camera of that era (D2X, D2S? I can't remember), in which the camera WB actually does affect the analog channel gains in the sensor and so WB does in fact affect RAW data - but that is an odd ball case.
Ken
-
Ray,
What might be confusing you is that there are a few photographers who will use external color correction filters to equalize the channel histograms and in this case your concerns are valid. What everyone here is talking about for UniWB is just setting up a WB in the camera to make the in camera histograms better reflect the RAW histograms. As noted, this doesn't change the RAW data at all. There is one notable exception, I believe it is the D200 or some other Nikon camera of that era (D2X, D2S? I can't remember), in which the camera WB actually does affect the analog channel gains in the sensor and so WB does in fact affect RAW data - but that is an odd ball case.
Ken
I see. In that case it's all just a 'strom in a tea cup' as far as I'm concerned. For a moment I thought there was a real possibility of extending the dynamic range of the camera, using UNIWB. Can't see what all the fuss is about.
-
I thought, by applying a UNIWB, the channel most susceptible to clipping would be brought back into line with the other channels, allowing for a greater exposure.
That happens only in the JPEG, and that is the principle why UniWB'ed histograms are (or should) be more reliable (and less pesimistic) than those obtained with any other WB. In a UniWB'ed JPEG histogram, R and B channels typically have less exposure than if they were calculated with another WB setting, precisely because any typical WB consists of increasing exposure in the R and B channels.
But the RAW data, and hence (never knew if this is the proper way to use the English word 'hence') what you can obtain in your RAW converter, is exactly the same no matter what WB you used. RAW data only depends on the triad: ISO, shutter and aperture; UniWB is just intended to assist you to be more precise in handling them.
Regards
-
That happens only in the JPEG, and that is the principle why UniWB'ed histograms are (or should) be more reliable (and less pesimistic) than those obtained with any other WB. In a UniWB'ed JPEG histogram, R and B channels typically have less exposure than if they were calculated with another WB setting, precisely because any typical WB consists of increasing exposure in the R and B channels.
But the RAW data, and hence (never knew if this is the proper way to use the English word 'hence') what you can obtain in your RAW converter, is exactly the same no matter what WB you used. RAW data only depends on the triad: ISO, shutter and aperture; UniWB is just intended to assist you to be more precise in handling them.
Regards
Now I understand why I didn't get involved in that 2008 thread on UNIWB you linked to, involving Panopeeper, John Sheehy and Jonathan Wienke. It didn't make sense. A lot of huffing and puffing about nothing. ;D
-
But knowing that the camera WB setting doesn't affect RAW data is basics Ray! that is first course ;D
-
But knowing that the camera WB setting doesn't affect RAW data is basics Ray! that is first course ;D
But setting the camera's jpeg to minimum contrast, minimum saturation and minimum sharpness provided a sufficiently accurate histogram for me. Your UNIWB may provide a slightly more accurate histogram, but with an extra degree of trouble that may not be warranted.
-
I'm now getting an isight into how some people view these incessant, obsessive, technical discussions over what may appear to them to be trivial concerns.
Many of us get excited about a new camera that provides an extra stop of DR or even a 25% increase in resolution, but I don't believe I've ever witnessed before on this site so much effort and weighty discussion over such a small advantage.
When I saw that 6 page thread on UNIWB, with such detailed technical posts from John Sheehy, Panopeeper and Johnathan Wienke, and links to images that could be downloaded to help one achieve an accurate UNIWB, I thought for a moment there was a real advantage being discussed, and I wondered if I should go to the trouble of using such a procedure myself, under the misapprehension I might eke out an extra 1/2 stop of DR, or so.
I should have realised immediately that changing the WB would not affect the RAW data, or allow for any increase in exposure before clipping took place.
You certainly fooled me, Guillermo. ;D
-
Quote
I should have realised immediately that changing the WB would not affect the RAW data, or allow for any increase in exposure before clipping took place.
Unquote
Surely this is semantics. If I import an image into ACR the image has setting > as shot. If I change it to daylight more often than not some clipping will appear shown by the red flashes. Now the RAW data will not be changed but they way I then process it will be changed because I will lower the exposure slider more than I would if I left it at as shot. I like to look at things from a practical sense rather than a purely theoretical one. It seems to me that a lot of the posters on here like to dance on a pin head for the sake of it and argue endlessly theoretical points for the sake of it. I sometimes wonder how good a photographer they are in reality when it comes to producing decent prints? ::)
-
Its not semantics as I see it at all. The WB has zero effect on the raw data. There is a suggestion that is completely separate from the raw, every converter will use it differently or maybe ignore it. You could have some separate metadata that again has no effect on the raw and a proprietary raw converter may use it and produce a visual result that is completely different from another raw converter. And yes, if that suggestion applies some rendering that affects the histogram or clipping, you’ll see it and if you blindly let that apply to the rendering, you’ll get that result. The raw however wasn’t altered a lick.
I should have realised immediately that changing the WB would not affect the RAW data, or allow for any increase in exposure before clipping took place.
Correct. What you do in any number of raw converters with any number of rendering controls at this point will of course.
-
"The WB has zero effect on the raw data."
It does have an impact on your decision making at time of capture, which is the whole point regarding UniWB.
-
It does have an impact on your decision making at time of capture, which is the whole point regarding UniWB.
How so?
-
Well, I'll kick off with a few methods that have worked for me.
(2) Buy a new camera which has a linear noise and dynamic range response (on the graphs at DXO Mark) that reduces by one f stop (or EV) for every doubling of ISO, so that there is no fundamental image quality advantage to increasing ISO.
The Nikon D7000 and Pentax K5 are in this category. With such a camera, you have only one worry regarding ETTR, and that's inadvertent overexposure at base ISO.
Unfortunately, the review image with this method can look even worse than the first method, if you've been underexposing at base ISO instead of using a higher ISO. It may be so dark you can hardly see it at all and you may have to flatly refuse to show it to anyone in order to protect your reputation.
??? It is not clear to me how this solution makes it easier to implement ettr. Even with cameras which do not have a linear noise response to increasing ISO, boosting the ISO (as opposed to increasing the exposure) may diminish read noise but is never a correct solution to achieving proper ettr. Using such a camera, you suggest that one need only worry about inadvertent overexposure at base ISO; how does such a camera prevent inadvertent underexposure at base (or any other) ISO? And finally, even if no one is looking over your shoulder, a very dark review image with any camera suggests that you are seriously underexposing or that your ISO is too low, or both.
-
??? It is not clear to me how this solution makes it easier to implement ettr. Even with cameras which do not have a linear noise response to increasing ISO, boosting the ISO (as opposed to increasing the exposure) may diminish read noise but is never a correct solution to achieving proper ettr. Using such a camera, you suggest that one need only worry about inadvertent overexposure at base ISO; how does such a camera prevent inadvertent underexposure at base (or any other) ISO? And finally, even if no one is looking over your shoulder, a very dark review image with any camera suggests that you are seriously underexposing or that your ISO is too low, or both.
It's always possible to make a mistake whatever the method used for achieving ETTR, even if one had a camera that presents a true RAW histogram. One could make a mistake thinking the camera was set to ISO 400 when in fact it was on ISO 100, or vice versa, simply because one was in a hurry to get the shot and not paying full attention.
I'm basically assuming that anyone who is so concerned about getting a really accurate ETTR, is likely to be using the camera in full manual mode, because getting an appropriate shutter speed and desired F stop is generally of more importance than getting a 'spot on' ETTR. At least it is for me.
If one accepts this as a reasonable premise, then an underexposure at base ISO would result only because the shutter speed selected for the conditions of the scene, to freeze subject movement and/or camera shake, were too fast in relation to the base ISO setting, but necessary regardless.
To use base ISO in these circumstances would be considered a mistake with certain models of cameras, such as all recent Canon DSLRs, because it's always better with such cameras, to raise ISO to achieve an ETTR in relation to that increase in ISO.
When thus raising ISO to avoid underexposure at base ISO, because of the image quality advantage of raising ISO, one still has the same ETTR problem of the possibility of blown highlights because of a slight miscalculation.
With cameras such as the D7000 and K5, there should be no such concern because an underexposure at base ISO provides the same fundamental image quality, as the same exposure with same settings of shutter speed and aperture, used at an oppropriately higher ISO. Got it?
-
With cameras such as the D7000 and K5, there should be no such concern because an underexposure at base ISO provides the same fundamental image quality, as the same exposure with same settings of shutter speed and aperture, used at an oppropriately higher ISO. Got it?
Finally bought such camera which is supposed to represent latest sensor technology,
and my 'concerns' soon started to shift from ETTR to 'single shot HDR'.
Peter
--
-
I see. In that case it's all just a 'strom in a tea cup' as far as I'm concerned. For a moment I thought there was a real possibility of extending the dynamic range of the camera, using UNIWB. Can't see what all the fuss is about.
The physical noise and clipping characteristics of the sensor can (of course) not be improved. But by helping you place the "signal" optimally compared to the (more or less) fixed noise and clipping point, you can get better DR using UniWB/ETTR/"insert good recording practice here" than not doing it. If e.g. 1 stop more DR is crucial to you, and worth the hazzle, I dont see the problem.
Automated exposure has some safeguards (probably for good reason), and in-camera histograms are flawed. By improving the feedback on signal level, you may buy yourself some DR headroom that otherwise would have to be bought in the for of a newer camera.
Of course, knowing the AE of your camera, its margins, and perhaps doing bracketing might in the end give similar results and less pain for some users?
-h
-
(2) Buy a new camera which has a linear noise and dynamic range response (on the graphs at DXO Mark) that reduces by one f stop (or EV) for every doubling of ISO, so that there is no fundamental image quality advantage to increasing ISO.
The Nikon D7000 and Pentax K5 are in this category. With such a camera, you have only one worry regarding ETTR, and that's inadvertent overexposure at base ISO.
I read the DxO curve as follows. There is a linear relationship between S/N and ISO. This means that exposing EV -2 at ISO 100 has the same noise as exposing EV 0 at ISO 400. In other words you can shoot at base ISO and increase exposure in post processing and get the same noise characteristics as if you had raised the ISO in the camera and exposed the same stops higher at a higher ISO with the risk of clipping highligts. However this is not the same noise as at base ISO!
Now as far as I would analyze this there are a couple of issues with this approach.
1) The noise will be higher when you underexpose and raise exposure in post processing just like if you shot a higher ISO. If you shoot landscape at is not a good idea from a noise perspective to underexpose by two stops relative to ETTR and the raise it by two stops in pp. Correctly exposed (ETTR) at base ISO will always give the lowest noise.
2) Underexposed images will loose gradations since there are fewer bits to represent the shadows. That's a reality of mathematics and nothing else. 14 bit RAW files helps on this, of course. Also deep shadows may not be so misrepresented with few gradations as they are deep shadows. But an extreme example of raising the exposure by 4 stops to go from ISO 100 to ISO 1600. The tonality will clearly be impacted as the shadows have now been lifted by 4 stops in pp. Btw. the max you can do in Lightroom and ACR without workarounds.
3) The DR is reduced by as many stops that the image is underexposed compared to perfect ETTR.
So just underexposing because it doesn't matter is simply not true. However these cameras with 14 stops of DR and a linear SN curve will do better than older ones when lifting shadows in a single exposure when lifting shadows. Like when you do a blend of multiple RAW conversions in e.g. Photomatix og you compact the DR in e.g. Lightroom using the tone curve or recovery and fill light combined with a suitable contrast curve (sort of like a tone mapping done in Lightroom). Alternatively using graduated filters and/or the brush to redistribute the light. In such cases a camera like the 1Ds mkIII can handle in my experience about a 2 stop lifting the shadows with no adverse effects. More than that you start seeing the extra noise and reduction of details when we compare to a clean ISO 100 shot of similar details.
-
2) Underexposed images will loose gradations since there are fewer bits to represent the shadows. That's a reality of mathematics and nothing else. 14 bit RAW files helps on this, of course. Also deep shadows may not be so misrepresented with few gradations as they are deep shadows. But an extreme example of raising the exposure by 4 stops to go from ISO 100 to ISO 1600. The tonality will clearly be impacted as the shadows have now been lifted by 4 stops in pp. Btw. the max you can do in Lightroom and ACR without workarounds.
Underexposing means losing gradations in the RAW histogram, but in practice you will _NOT SEE_ that loss because noise dithers posterization. You will not see even in the output histogram after RAW conversion since after demosaicing and output colour profiling, the 16-bit histogram gets full of levels everywhere. Just underexpose some RAW, lift the shadows, and look for posterization.
All this story about ETTR being helpful in having more tonal levels and prevent posterization always extends like a virus! (without anyone providing a single evidence of it).
ACR allows to lift exposure in the shadows (which is where the discusion about gradations and ETTR makes sense) by 7 stops: 4 stops through the Exposure slider + 3 stops by setting Brightness to +150.
-
So you agree with my points, except that you argue that the degradation in gradations will be hard to see.
As far as I'm aware the exposure slider and the brightness slider in Lightroom are not the same. Exposure sets the white point and brightness moves midtones without moving the white point. I think Schewe can comment on this.
-
So you agree with my points, except that you argue that the degradation in gradations will be hard to see.
As far as I'm aware the exposure slider and the brightness slider in Lightroom are not the same. Exposure sets the white point and brightness moves midtones without moving the white point. I think Schewe can comment on this.
No, I don't agree. It will not be hard to see, it is impossible to see the gradations, so talking about losing gradations is misleading because you can conclude that there is some advantage about gradations thanks to ETTR. I am ready to change my mind if you provide an example of those gradations.
Gradations become usually visible in areas with plain colours and very low noise, for example when converting skies to 8-bit JPEG. So paradoxically, ETTR will help in making banding appear in the highlight areas without textures. An ETTR'ed sky is more prone to banding than an underexposed sky.
The Brightness slider in ACR is the same as Exposure for the shadows, that means shadows can be lifted by 7 stops making both sliders work together. The straight lines are Exposure settings, the curves are Brightness. In the shadows they are the same; for instance Bright +50 will lift the shadows by 1EV:
(http://www.guillermoluijk.com/article/acrps/exp_brillo_acr.gif)
-
You can look back to see what the other points were on 1) and 3). I take that you agree on these points.
-
You can look back to see what the other points were on 1) and 3). I take that you agree on these points.
I agree with 1 (noise is the one and only reason for ETTR), but disagree with 3.
Not all cameras loose 1 full stop of DR when underexposing by 1 stop, that depends on the dB/EV slope of the SNR response of the particular sensor. Canons typically have 6dB/EV in the shadows because they are read noise limited sensors, so a Canon will loose 1 full stop of DR for every 1 stop of underexposure. The late Pentax K5 and Nikon D7000 sensor however have softer slopes, closer to 3dB/EV for being rather photon noise limited sensors, so one of these cameras will only loose 0,5 stops of DR for every 1 stop of underexposure. The conclusion is that although ETTR is still very recommended on these late cameras to improve SNR, it is not so critical as in Canons.
The plots may me somewhat difficult to interpret, but they have the key to all this:
(http://www.guillermoluijk.com/tutorial/noisedr/curvassnrnorm.gif)
Regards
-
Thanks, so we agree on point no. 1 and no. 3 partially with the comments you made here. Now, how did you arrive at the curves you made in your last response?
-
Thanks, so we agree on point no. 1 and no. 3 partially with the comments you made here. Now, how did you arrive at the curves you made in your last response?
By measuring pairs (RAW exposure, SNR) and plotting them. I did it over patches of uniform colour, the procedure is explained here: NOISE AND DYNAMIC RANGE MEASUREMENT. CANON 5D vs 5D2 vs 7D vs PENTAX K5 (http://www.guillermoluijk.com/tutorial/noisedr/index.htm) (Spanish, hope the online translation suffices).
-
There is another very effective way of comparing an ETTR at, say, ISO 1600, with the same exposure at ISO 100, and that is to take the shots of the same real world, 'high DR' scene, and then compare results, after your best processing effort in your converter of choice.
-
A couple of notes:
- ETTR is about maximizing signal-to-noise. That's all. In-camera histograms and white balance settings don't help, because they don't reflect how the raw data is recorded. It is true that photographers can try to use these tools to ETTR, but unfortunately these tools are misleading for this purpose, and can actually lead photographers down the wrong path (see the red flower example in Michael's article).
- ACR's Exposure and Brightness are essentially the same, except for the handling of highlights. Exposure will hard-clip the highlights, whereas Brightness will not. For example, if you push Exposure to +1, then anything that was in the top stop will get clipped off -- gone -- similar to digital exposure. In contrast, if you push Brightness from +50 to +100, then anything that was in the top stop will get smoothly rolled off (instead of clipped off) -- similar to film.
-
A couple of notes:
- ETTR is about maximizing signal-to-noise. That's all.
Halleluiah, we can finally put this myth about the number of levels in the highlights to rest. If Michal were to remove that concept from his otherwise fine post, I would be in complete agreement. My opinion may not carry much weight, but the opinions of experts like yourself and others do.
- ACR's Exposure and Brightness are essentially the same, except for the handling of highlights. Exposure will hard-clip the highlights, whereas Brightness will not. For example, if you push Exposure to +1, then anything that was in the top stop will get clipped off -- gone -- similar to digital exposure. In contrast, if you push Brightness from +50 to +100, then anything that was in the top stop will get smoothly rolled off (instead of clipped off) -- similar to film.
An excellent point!
Regards,
Bill
-
Halleluiah, we can finally put this myth about the number of levels in the highlights to rest.
Amen, brother
Sandy
-
A couple of notes:
- ETTR is about maximizing signal-to-noise. That's all. In-camera histograms and white balance settings don't help, because they don't reflect how the raw data is recorded. It is true that photographers can try to use these tools to ETTR, but unfortunately these tools are misleading for this purpose, and can actually lead photographers down the wrong path (see the red flower example in Michael's article).
Yes, maximizing signal to noise. And underexposing many stops and raise in post will reduce levels. That's a fact, but how many we will see in a print is another question. Underexposing also reduces DR as mentioned although the number of stops depends on the camera.
- ACR's Exposure and Brightness are essentially the same, except for the handling of highlights. Exposure will hard-clip the highlights, whereas Brightness will not. For example, if you push Exposure to +1, then anything that was in the top stop will get clipped off -- gone -- similar to digital exposure. In contrast, if you push Brightness from +50 to +100, then anything that was in the top stop will get smoothly rolled off (instead of clipped off) -- similar to film.
Exposure sets the white point and the black point and brigtness moves midtones between those points. So they are not the same.
-
Exposure sets the white point and the black point and brigtness moves midtones between those points. So they are not the same.
Exposure sets the white point, and secondary changes may be seen in the black point. The black point is set by the blacks control. Obviously they are not exactly the same, or both controls would not be needed. If I were you, I would be careful about preaching to Eric Chan, since he is a senior engineer on the ACR team and he knows what he is talking about.
Regards,
Bill
-
underexposing many stops and raise in post will reduce levels. That's a fact, but how many we will see in a print is another question.
Probably Chuck Norris is the only one who will.
-
Exposure sets the white point, and secondary changes may be seen in the black point. The black point is set by the blacks control. Obviously they are not exactly the same, or both controls would not be needed. If I were you, I would be careful about preaching to Eric Chan, since he is a senior engineer on the ACR team and he knows what he is talking about.
Regards,
Bill
Yes, black point is set by the blacks slider, I'm sorry that fell out of my post. I'm not preaching to anybody ;) I'm not aware of what position Mr. Chen has. I would certainly welcome an explanatory statement from him about the differences some more detail than posted here. I have used Lightroom and ACR quite a lot and it is obvious that exposure and brightness is not the same. I also have the videos from Schewe and Michael and they explain the differences between exposure and brightness in Lightroom. So if this is wrong or I'm wrong I would welcome being corrected ;)
-
Hans,
Go back and read what Eric wrote: "ACR's Exposure and Brightness are essentially the same, except for the handling of highlights. Exposure will hard-clip the highlights, whereas Brightness will not. For example, if you push Exposure to +1, then anything that was in the top stop will get clipped off -- gone -- similar to digital exposure. In contrast, if you push Brightness from +50 to +100, then anything that was in the top stop will get smoothly rolled off (instead of clipped off) -- similar to film."
The big difference between Exposure and Brightness is the fact that + Brightness gets rolled off so it won't clip while + Exposure is designed to clip. There are differences between -Brightness and -Exposure as well, -Exposure has a degree of highlight clipping reduction while -Brightness shifts the midtone down more.
As far as Eric giving you more...well he's one of the top ACR/LR engineers working directly with Thomas Knoll. Exactly how much he can say about anything can only be determined by him. He also indicated somewhere on one of the many (too many) threads about ETTR that you can get more info on how ACR/LR works by downloading and reading the DNG SDK available here (http://www.adobe.com/products/photoshop/extend.displayTab2.html).
-
ETTR is about maximizing signal-to-noise. That's all.
Halleluiah, we can finally put this myth about the number of levels in the highlights to rest. If Michal were to remove that concept from his otherwise fine post, I would be in complete agreement. My opinion may not carry much weight, but the opinions of experts like yourself and others do.
An excellent point!
Regards,
Bill
I think we're engaging in more than semantics here, Bill. One can't disagree with Eric that ETTR is about maximizing signal-to-noise, but the 'That's all' bit is misleading in my view.
A more precise definition would be, "Setting the exposure to maximise signal-to-noise whilst simultaneously retaining full detail in all highlights in the scene that one considers important."
The process of merging to HDR is also all about maximising signal-to-noise, and often one of the 3 or 5 or 7 or 9 different exposures taken for that purpose, will likely be an ETTR. However, the exposure that maximises signal-to-noise will be an obvious overexposure.
It's very easy to maximise signal-to-noise. Just overexpose, and the more overexposure the better, as regards signal-to-noise. From a strictly logical perspective, all overexposures could be considered as ETTRs, if you wish to add 'That's all'.
-
It's very easy to maximise signal-to-noise. Just overexpose, and the more overexposure the better, as regards signal-to-noise. From a strictly logical perspective, all overexposures could be considered as ETTRs, if you wish to add 'That's all'.
You could consider clipping to be some kind of noise (undesired modification of the image as projected onto the sensor). But your point still stands: if artistic goals (DOF, motion blur) allows, you should expose so hot as to maximize signal level compared to noise level, while avoiding exposing so hot as to clip bright parts of the scene that you want accurately captured.
-h
-
The process of merging to HDR is also all about maximising signal-to-noise, and often one of the 3 or 5 or 7 or 9 different exposures taken for that purpose, will likely be an ETTR. However, the exposure that maximises signal-to-noise will be an obvious overexposure.
It's very easy to maximise signal-to-noise. Just overexpose, and the more overexposure the better, as regards signal-to-noise. From a strictly logical perspective, all overexposures could be considered as ETTRs, if you wish to add 'That's all'.
That depends on how you define signal. Shown below are results from the Nikon D3 with duplicate flat fields of the green1 channel taken for sensor analysis. The shot noise is determined by subtracting the two flat fields after adding an offset to one field to prevent negative values. The resultant value is the shot noise for two frames, so one divides by the square root of 2 to obtain the noise for one image. Sigma (standard deviation) is shown for the individual values for each pair and Sigma 3 is the standard deviation for one frame. One notes that the noise increases as one goes from frames 55 to 39 and the SNR decreases as expected. However, when the sensor saturates, the noise is clipped and the SNR reached very high values. However, this is not useful SNR, since the signal has been clipped. The same trend applies for total noise rather than shot noise.
In ETTR exposing, when the DR of the scene is greater than that of the sensor, SNR will increase with increasing exposure until the sensor saturates. Further exposure will clip the highlights but the DR will remain constant with further exposure as one stop of shadows is gained for each loss of a stop in highlights. The SNR of the overexposed shadow values will be improved. This is the basis for HDR photography.
Regards,
Bill
-
A more precise definition would be, "Setting the exposure to maximise signal-to-noise whilst simultaneously retaining full detail in all highlights in the scene that one considers important."
The process of merging to HDR is also all about maximising signal-to-noise, and often one of the 3 or 5 or 7 or 9 different exposures taken for that purpose, will likely be an ETTR. However, the exposure that maximises signal-to-noise will be an obvious overexposure.
It's very easy to maximise signal-to-noise. Just overexpose, and the more overexposure the better, as regards signal-to-noise. From a strictly logical perspective, all overexposures could be considered as ETTRs, if you wish to add 'That's all'.
Yes, but I don't consider a clipped (saturated) signal to be useful. ;)
It is up to the photographer to decide what is "important" to retain in the highlights. For example, it may be fine (even desirable) to clip some specular reflections to pure white -- no detail recorded in those areas. In other cases, it may be desirable to keep the full range of highlights. This may make it difficult or impossible to increase the exposure further.
So, to clarify: the motivation behind ETTR is all about maximizing signal-to-noise, but in the field ETTR cannot always be used for all situations.
-
Yes, but I don't consider a clipped (saturated) signal to be useful. ;)
It is up to the photographer to decide what is "important" to retain in the highlights. For example, it may be fine (even desirable) to clip some specular reflections to pure white -- no detail recorded in those areas. In other cases, it may be desirable to keep the full range of highlights. This may make it difficult or impossible to increase the exposure further.
So, to clarify: the motivation behind ETTR is all about maximizing signal-to-noise, but in the field ETTR cannot always be used for all situations.
We seem to be confusing the 'objective' with the 'subjective', which is why I would consider my definition to be more precise because it includes both the objective and the subjective, ie. ""Setting the exposure to maximise signal-to-noise whilst simultaneously retaining full detail in all highlights in the scene that one considers important."
What one considers important is a subjective assessment. Setting exposure to maximise the total number of levels captured in the scene could be considered as the basic, objective definition of ETTR, which is a good starting point.
The obvious example, which I've used before somewhere, is a scene of wildlife in the shade of a tree with significant areas of sky visible through the branches. One photographer thinks the sky (its detail or its color) is sufficiently interesting to retain, and increases his shutter speed accordingly. Another photographer decides to use a slower shutter speed to reduce noise in the main subject which he considers is the bird or wildlife, at the risk perhaps of failing to freeze movement.
It makes little sense from an objective perspective to describe both exposures, which may differ significantly, an ETTR. From an objective perspective, the shot that retains detail in the sky, to point of near clipping, is an ETTR. The other shot is a deliberate overexposure for esthetic or subjective reasons.
-
Maybe we should officially christen the term "Subjective ETTR". But, this a purely subjective suggestion....or is it? :o
-
Maybe we should officially christen the term "Subjective ETTR". But, this a purely subjective suggestion....or is it? :o
Ha! The lack of understanding of the difference between the subjective and the objective is perhaps the major contributor to all the troubles in the world.
I would prefer it if the definition of ETTR (Extra Terrestrial Tryannosaurus Rex) would confine itself to objective facts, ie, a full exposure short of clipping the highlights, excluding specral highlights which by definition are very intense and cover a very small area of the frame.
Why complicate things with a multitude of different subjective assessments which can have no ultimate resolution.
If you want to blow highlights, such as a sky, because it's of less importance than the rest of the image, then let's call it an 'overexposure for esthetic reasons'.
-
What one considers important is a subjective assessment. Setting exposure to maximise the total number of levels captured in the scene could be considered as the basic, objective definition of ETTR, which is a good starting point.
The old myth regarding the number of levels dies hard. As Eric and Emil Martinec have pointed out it is the SNR that benefits from ETTR. Due to dithering by noise, the highlights have many fewer levels than presumed from a purely mathematical analysis, but also do not show posterization because of the noise.
The obvious example, which I've used before somewhere, is a scene of wildlife in the shade of a tree with significant areas of sky visible through the branches. One photographer thinks the sky (its detail or its color) is sufficiently interesting to retain, and increases his shutter speed accordingly. Another photographer decides to use a slower shutter speed to reduce noise in the main subject which he considers is the bird or wildlife, at the risk perhaps of failing to freeze movement.
It makes little sense from an objective perspective to describe both exposures, which may differ significantly, an ETTR. From an objective perspective, the shot that retains detail in the sky, to point of near clipping, is an ETTR. The other shot is a deliberate overexposure for esthetic or subjective reasons.
Since the writings of Bruce Fraser, the rationale of ETTR has always been to place highlights in which one wishes to preserve image detail near the clipping point. If you want to preserve the sky in your situation, you place the sky highlights near clipping. If shadow detail is more important and the DR of the scene exceeds that of the camera, then one will have to allow the highlights to clip or else take 2 or more exposures. The principles of ETTR remain the same, but subjective decisions govern the exposure. In both cases one gives maximal exposure to achieve the intended effect.
Regards,
Bill
-
Since the writings of Bruce Fraser, the rationale of ETTR has always been to place highlights in which one wishes to preserve image detail near the clipping point. If you want to preserve the sky in your situation, you place the sky highlights near clipping. If shadow detail is more important and the DR of the scene exceeds that of the camera, then one will have to allow the highlights to clip or else take 2 or more exposures. The principles of ETTR remain the same, but subjective decisions govern the exposure. In both cases one gives maximal exposure to achieve the intended effect.
Regards,
Bill
So what's new? What's all the fuss about?
-
So what's new? What's all the fuss about?
Good question. Exposing "hot" is common sense for anyone doing digital recordings of any kind of data. For people that have only used film cameras, this could be new. For people relying on automatic camera exposure algorithms, or in-camera histograms, knowledge about the impresise feedback and built-in margins could be new.
-h
-
While the basic concept of ETTR is objective, my view is that the practical application of ETTR is nearly always subjective to some degree. In many common situations, some pixels will get clipped regardless of the choice of exposure (e.g., bright light source in the scene). To be effective, a camera algorithm that implemented ETTR would need to have some algorithm (or possibly user control) to decide the clip threshold, e.g., the difference between clipping .01% and 1% can be huge in some scenes.
-
While the basic concept of ETTR is objective, my view is that the practical application of ETTR is nearly always subjective to some degree. In many common situations, some pixels will get clipped regardless of the choice of exposure (e.g., bright light source in the scene). To be effective, a camera algorithm that implemented ETTR would need to have some algorithm (or possibly user control) to decide the clip threshold, e.g., the difference between clipping .01% and 1% can be huge in some scenes.
Agree with the differences that would occur; I just want us to have more options that we have right now. I'm reading up on uniWB and maybe it's a decent solution but it does require some work to implement it (just as testing your metering system and JPG histogram to calibrate that for the real world).
-
The old myth regarding the number of levels dies hard. As Eric and Emil Martinec have pointed out it is the SNR that benefits from ETTR. Due to dithering by noise, the highlights have many fewer levels than presumed from a purely mathematical analysis, but also do not show posterization because of the noise.
Bill
Really! Do you mean it's possible to have an unprocessed ettr RAW file with the resultant benefit of a better SNR but without the concomitant increase in the number of levels that one expects from an ETTR? I never realised that.
I always had some vague notion that an ETTR exposure, as compared to an underexposure, would allow a greater number of levels to be recorded.
Of course, I understand whether or not such increased levels can be seen is another issue. In the midtones and shadows, they no doubt can be seen. In the upper tones, many of the levels may to be too similar for the eye to distinguish.
Is this a myth then?
-
While the basic concept of ETTR is objective, my view is that the practical application of ETTR is nearly always subjective to some degree. In many common situations, some pixels will get clipped regardless of the choice of exposure (e.g., bright light source in the scene). To be effective, a camera algorithm that implemented ETTR would need to have some algorithm (or possibly user control) to decide the clip threshold, e.g., the difference between clipping .01% and 1% can be huge in some scenes.
Indeed! And herein lies the problem; deciding how much clipping is best. An automatic exposure which is accurate from an ETTR perspective, resulting in a RAW histogram on the review screen that takes the place of a UNIWB adjustment, whilst simultaneously providing a realistic and natural rendition of the picture taken, would be of some benefit.
But as soon as we get back to user-controlled adjustments made on the basis of subjective assessments of the brightness characteristics of the scene, such as trying to assess whether the area of sky visible through the branches of a tree constitutes 1% or 5% or 10% of the image, we are likely to fail to capture the moment through too much stuffing around.
One might as well just use the spot-meter technique directed at the brightest part of the scene one wants to preserve, then make the appropriate increase in exposure.
-
Really! Do you mean it's possible to have an unprocessed ettr RAW file with the resultant benefit of a better SNR but without the concomitant increase in the number of levels that one expects from an ETTR? I never realised that.
The number of levels is fixed upon quantization by the ADC. The lower exposure levels may suffer from noise to the point that they are hard to discriminate in a single exposure. It's solely the improved S/N ratio that will seemingly add levels, but that is just the result of the improved S/N ratio. The levels were there all the time, it's just that some may not be very useful.
I always had some vague notion that an ETTR exposure, as compared to an underexposure, would allow a greater number of levels to be recorded.
They are recorded anyway, but may look swamped by noise, especially when underexposed. Improving the Poisson noise statistics by increasing the photon count is all it takes to improve the discrimination. Increasing the exposure time without clipping relevant highlights is one method to achieve that, stacking and averaging multiple exposures of scenes without moving objects is another. It's all about improving the photon statistics.
Cheers,
Bart
-
The number of levels is fixed upon quantization by the ADC. The lower exposure levels may suffer from noise to the point that they are hard to discriminate in a single exposure. It's solely the improved S/N ratio that will seemingly add levels, but that is just the result of the improved S/N ratio. The levels were there all the time, it's just that some may not be very useful.
They are recorded anyway, but may look swamped by noise, especially when underexposed. Improving the Poisson noise statistics by increasing the photon count is all it takes to improve the discrimination. Increasing the exposure time without clipping relevant highlights is one method to achieve that, stacking and averaging multiple exposures of scenes without moving objects is another. It's all about improving the photon statistics.
Cheers,
Bart
A wonderful exercise in obfuscation, Bart. Can I nominate you for a prize? ;D
The question is simple. Does an increase in exposure, up to the point of clipping of highlights, result in an increase in the number of levels recorded? Yes, No or Maybe?
If maybe, then please decribe the circumstances.
-
The question is simple. Does an increase in exposure, up to the point of clipping of highlights, result in an increase in the number of levels recorded? Yes, No or Maybe?
If maybe, then please decribe the circumstances.
As people have been saying (repeatedly) the issue is S/N ratio, not the number of levels. At a given illumination level (number of photons = S), the noise goes as sqrt(S), so the S/N = sqrt(S). You improve S/N by capturing more photons -- raising the exposure. Regardless how many levels there are, two tones are distinguishable (on average) if they are spaced more than the noise. So there are roughly sqrt(S2)-sqrt(S1) distinguishable tones in the exposure range between S1 and S2. If you increase the exposure by a stop, you increase this number of distinguishable tones by a factor sqrt(2)~1.4, even though the number of levels in the raw data has doubled. Similarly, if you took the same picture with a D700 in 12-bit mode and 14-bit mode, the 14-bit capture would have 4x more levels in the raw data, but the same number of distinguishable tones, because at the same exposure the same number of photons is captured. It's the S/N, not the number of levels.
There are corrections to the above analysis due to read noise, that are only important at low exposure (deep, deep in the shadows at low ISO, creeping upward into higher zones as the ISO is raised).
-
As people have been saying (repeatedly) the issue is S/N ratio, not the number of levels.
Indeed and a good point to reinforce. But if we look at the distribution of levels in a linear encoded file, we see half of all that data in the first stop of highlight and on the other end, the smallest number of levels. The question is, how do we describe in a sentence or two the relationship of those fewer levels in the last stop, the noise that results there with a higher or lower S/N ratio? In the classic example shown on this site and originally by Bruce Fraser, the last stop of shadow detail had 16 levels (as opposed to the first at 2048). Its a good point that no matter the exposure, its still 16 levels, how do we define the relationship and effect of these fewer levels, noise and S/N?
A wonderful exercise in obfuscation, Bart. Can I nominate you for a prize?
In the original ETTR article here, Ian Lyons is quoted as saying: The ideal exposure ensures that you have maximum number of levels describing your image without loosing important detail in the highlights. The closer you get to this ideal then the more of those levels are being used to describe your shadows. How about working on a sentence that takes all of the above into account and explains the relationship and effect of these fewer levels, noise and S/N.
-
As people have been saying (repeatedly) the issue is S/N ratio, not the number of levels. ...
If you increase the exposure by a stop, you increase this number of distinguishable tones by a factor sqrt(2)~1.4, even though the number of levels in the raw data has doubled. Similarly, if you took the same picture with a D700 in 12-bit mode and 14-bit mode, the 14-bit capture would have 4x more levels in the raw data, but the same number of distinguishable tones, because at the same exposure the same number of photons is captured. It's the S/N, not the number of levels.
Data from DXO can be used to confirm Emil's analysis. I will use the Phase One P65+ as an example, since it is generally regarded as one of the highest performers. The tonal range is the number of discrete gray levels that can be discerned in the image. From the DXO graph and derived table, one can see that the tonal range decreases by 0.5 bits for each doubling of ISO. Doubling of ISO decreases the number of photons by a factor of 2 and the number of levels by sqrt(2). At base ISO the P65+ has a tonal range of 8.64 bits or 399 levels.
The difference in the number of levels between ISO of 100 and ISO 50 reveals by subtraction that the brightest f/stop used for the base ISO of 50 contains 117 levels, far fewer than the 33,000 levels in the brightest f/stop of the 16 bit capture.
Regards,
Bill
-
In the classic example shown on this site and originally by Bruce Fraser, the last stop of shadow detail had 16 levels (as opposed to the first at 2048).
The fallacy of this reasoning is that the brightest f/stop of a 12 bit capture from a real camera contains far fewer than the claimed 2048 levels. See my previous post. Even the P65+ using a 16 bit ADC yields only 117 levels in the brightest f/stop.
Regards,
Bill
-
Really! Do you mean it's possible to have an unprocessed ettr RAW file with the resultant benefit of a better SNR but without the concomitant increase in the number of levels that one expects from an ETTR? I never realised that.
Yes. This is in fact exactly what happens when using ETTR with a Nikon camera that uses "lossy" compressed NEF format. ETTR doesn't get you any more levels because of the compressed NEF format and yet you still get all the benefits of ETTR since the SNR has improved. ETTR has never had a dang thing to do with "levels". It has always been about SNR and photon counting statistics despite what well meaning and capable photographers might write on their web sites.
Ken
-
Thanks for pointing this out. Honestly I never combined the DxO tonal measurements with ETTR and probably should have ;)
So even with a lot of underexposure like 5 stops on a 14 bit camera there is still bits enough to record the levels in the highlights og that this camera can capture at base ISO.
So thanks to those who pointed this out.
So ETTR is
1) Maximizing S/N ratio
2) Maximizing DR (some cameras more than others)
3) Levels which seems to almost drop off the list based on this discussion.
-
Its a good point that no matter the exposure, its still 16 levels, how do we define the relationship and effect of these fewer levels, noise and S/N?
How about working on a sentence that takes all of the above into account and explains the relationship and effect of these fewer levels, noise and S/N.
The sentence is pretty easy actually:
The number "levels" in the RAW file have no measurable effect on the noise at all from shadows to highlights, read noise and photon shot noise always are larger.
As an example, lets take the lowest read noise sensor in existence (for photography).
The D7000 sensor has a read noise of 1.4 times the least significant bit at base ISO. That means even ignoring photon shot noise at every exposure level read noise is randomly toggling more than one "level". Adding more levels anywhere on the scale from shadows to highlights will not improve noise.
Looked at another way, each "level" of the D7000 14-bit RAW file is equivalent to to just 2.6 photons hitting the pixel. Forget even the "16 levels" shadows. Lets look at just the bottom two bits "four levels" shadows. As stated the read noise is already greater than a single "level" but lets pretend it was made zero. At this "fourth level", two stops below the example you cite, we are accumulating about 10 photons. The shot noise from the counting statistics will be about 3.2 photons. Again, larger than a single "level".
At the risk of getting overly emphatic - levels don't matter - never have, never will. Photon counting statistics and read noise matter - always have, always will. If some idiot of an engineer were to ever design a camera with an ADC that caused the IQ to be limited by the number of levels they would be summarily fired.
Ken
-
Really! Do you mean it's possible to have an unprocessed ettr RAW file with the resultant benefit of a better SNR but without the concomitant increase in the number of levels that one expects from an ETTR? I never realised that.
I always had some vague notion that an ETTR exposure, as compared to an underexposure, would allow a greater number of levels to be recorded.
As the DXO data and Emil's analysis show, each stop of additional exposure increases the SNR by a factor of sqrt(2) and the number of distinguishable levels by the same factor. The number of levels is not doubled as the usual ETTR analysis would indicate.
Of course, I understand whether or not such increased levels can be seen is another issue. In the midtones and shadows, they no doubt can be seen. In the upper tones, many of the levels may to be too similar for the eye to distinguish.
Is this a myth then?
The question here is whether your image is constrained by SNR or the number of levels. The effect of SNR is obvious. An insufficient number of levels will result in posterization. As Emil has pointed out, the raw file is never posterized, but posterization can result from overzealous editing of the raw file. However, posterization is mitigated by dithering of the image by noise. In my work, I rarely see posterization with underexposed raw files. What about you?
The number of levels that the human visual system can distinguish is described by the Weber-Fechner law. Humans can distinguish a 1% difference in luminance, which corresponds to about 70 levels (see Norman Koren (http://www.normankoren.com/digital_tonality.html)). Fewer levels can be distinguished in the shadows. Excess numbers of levels in the brightest portions of the image can be discarded without affecting image quality.
Some subjective degradation in tonal quality may occur before overt posterizatioin, but this would be difficult to quantify.
Regards,
Bill
-
Indeed and a good point to reinforce. But if we look at the distribution of levels in a linear encoded file, we see half of all that data in the first stop of highlight and on the other end, the smallest number of levels. The question is, how do we describe in a sentence or two the relationship of those fewer levels in the last stop, the noise that results there with a higher or lower S/N ratio? In the classic example shown on this site and originally by Bruce Fraser, the last stop of shadow detail had 16 levels (as opposed to the first at 2048). Its a good point that no matter the exposure, its still 16 levels, how do we define the relationship and effect of these fewer levels, noise and S/N?
Maximizing the number of levels, and increasing S/N, often point in the same direction -- both go up as the exposure increases. But the details differ, and those differences can be important in some situations. Here is the base ISO number of distinguishable tones per channel, based on S/N, for the P65+ (continuing Bill's example), together with the number of raw levels (the calculated values include the effects of read noise as well as photon count fluctuations, aka "photon noise"):
top stop = 166 distinguishable tones, 32768 levels
2nd stop= 117 distinguishable tones, 16384 levels
3rd stop= 82 distinguishable tones, 16536 levels
4th stop= 57 distinguishable tones, 8192 levels
5th stop= 39 distinguishable tones, 4096 levels
6th stop= 26 distinguishable tones, 2048 levels
7th stop= 17 distinguishable tones, 1024 levels
8th stop= 10 distinguishable tones, 512 levels
9th stop= 6 distinguishable tones, 256 levels
10th stop= 3 distinguishable tones, 128 levels
11th stop= 2 distinguishable tones, 64 levels
12th stop= 0.9 distinguishable tones, 32 levels
So yes, they both trend upward as one increases the exposure; both imply ETTR for fixed ISO. But if one really subscribed to the "it's the number of levels that's important" mantra, one would be led to incorrect conclusions. For instance, suppose your base ISO exposure doesn't reach the top stop; does one do better by increasing the exposure a stop (eg by doubling the exposure time), or by doubling the ISO? If you double the ISO, you don't change the S/N because S/N on this camera is entirely determined by exposure. You do however double the number of levels, because the histogram is pushed to the right where the levels are denser. On the other hand, if you double the exposure, you again double the number of levels taken by a given patch of the image; you also increase the number of distinguishable tones by a factor ~1.4 according to the above. So the "#levels mantra" leads you to think you do just as well by raising the ISO at fixed exposure, whereas you actually only do better by increasing the exposure.
As a side note, on a camera such as the P65+, raising the ISO never increases the number of distinguishable levels; as far as quality of the raw data goes, raising the ISO accomplishes nothing but reduces the available highlight headroom (and therefore the room for increasing exposure, should you be able to do so without compromising DoF or motion blur requirements). (Caveat: raw data is no better, but your raw converter may treat it differently, since profiles often read the ISO and change the conversion accordingly, eg by applying more noise reduction at higher ISO.)
It is really quite remarkable how few distinguishable tones there really are in an image file (note that the above is per channel; modulo subtleties having to do with white balance, color filter response and output color gamut, the number of distinguishable colors is roughly the cube of the above numbers). The vast majority of all those wonderful raw values in the upper zones are wasted in quantizing noise, and not adding anything to image quality.
In the original ETTR article here, Ian Lyons is quoted as saying: The ideal exposure ensures that you have maximum number of levels describing your image without loosing important detail in the highlights. The closer you get to this ideal then the more of those levels are being used to describe your shadows. How about working on a sentence that takes all of the above into account and explains the relationship and effect of these fewer levels, noise and S/N.
Replace "levels" by "distinguishable tones" and you have a correct statement. Roughly, two tones S1 and S2 are distinguishable if |S2-S1|>N where the noise is N. If S1 and S2 differ by less than the noise, you don't know whether they are different tones, or the same tone pushed to different values by noise fluctuations.
-
Replace "levels" by "distinguishable tones" and you have a correct statement. Roughly, two tones S1 and S2 are distinguishable if |S2-S1|>N where the noise is N. If S1 and S2 differ by less than the noise, you don't know whether they are different tones, or the same tone pushed to different values by noise fluctuations.
Good, excellent. Assuming everyone is in agreement <g>
-
Good, excellent. Assuming everyone is in agreement <g>
Agreed, excellent and elegant explanation (As any other explanation from Emil, I would say).
-
What is the appropriate mantra? I would prefer "Maximize Exposure"; maximize subject to three constraints:
(1) maintaining needed DoF, which limits how much you can open up the aperture;
(2) freezing motion, which limits the exposure time;
(3) retaining highlight detail, by not clipping wanted highlight areas in any channel.
Note that ISO is not part of exposure. Exposure has only to do with aperture and shutter speed. Maximizing exposure guarantees that one captures as many photons as possible subject to photographic constraints, and therefore optimizes S/N.
How does ISO enter? It enters as a subsidiary aspect of optimizing S/N. On many cameras (those with CCD sensors, and the newer Sony Exmor sensors), there is little or no advantage to raising the ISO, which aids point (3) -- leaving the ISO at a low value may leave the histogram "to the left" for your chosen exposure, it will give more highlight headroom but will not degrade S/N; such cameras can safely be operated at close to their lowest ISO (the precise optimal ISO depends on the details of a given camera design). On the other hand, for many other CMOS sensor'd cameras, such as Canon's offerings, and Nikons with Nikon-designed CMOS sensors (D3/D700/D3s, for example), noise relative to exposure is improved by increasing the ISO; after you have maximized the exposure (ie by satisfying criteria (1) and (2)), you have a tradeoff to make for (3) -- raising the ISO lowers shadow noise (up to a camera-specific point of diminishing returns, usually about ISO 1600), therefore improving S/N, but reduces highlight headroom for your chosen exposure, so one has to decide how high the ISO can go and still keep wanted highlights unclipped.
Anyway, the prescription is to set the exposure (shutter speed and aperture only) according to (1) and (2); back off the exposure if at base ISO and you are compromising (3). If you are compromising (3) with your chosen exposure and you are not at base ISO, then you should have started with a lower ISO. Afterward, depending on the specifics of the camera's noise profile, further optimization results from raising the ISO, up to the limit specified by (3), or the camera's ISO point of diminishing returns, whichever is arrived at first.
So, it's (almost) all about ME. ;D
-
So, Emil...
If I correctly read your post....CCD and Exmor sensors, ETTR provides no benefit, is that correct?
On Canon and Nikon-designed CMOS sensors, is ETTR a benefit...up to the camera specific ISO limit?
John
-
So, Emil...
If I correctly read your post....CCD and Exmor sensors, ETTR provides no benefit, is that correct?
On Canon and Nikon-designed CMOS sensors, is ETTR a benefit...up to the camera specific ISO limit?
John
Rather, CCD and Exmor sensors provide little or no benefit to raising the ISO; Canon and Nikon-designed CMOS do provide benefit, up to a camera-specific ISO. You always get a benefit from increasing the exposure, if that does not conflict with your DoF and motion blur requirements.
I'd like to decouple the discussion from the histogram, which is where 'To The Right' refers to in ETTR. The most important consideration is always collecting the most photons during the exposure; you do that by maximizing the exposure (widest possible aperture consistent with DoF requirements, longest shutter time consistent with motion blur restrictions). Keeping a fixed (maximum subject to the constraints) exposure, where the histogram lies varies with ISO. You want the histogram as far to the left as possible to capture the most highlights, which you accomplish by lowering the ISO; you want to move it as far to the right as possible if in so doing you lower the camera's ISO-dependent noise contribution (the so-called read noise), which you can sometimes but not always be accomplished by raising the ISO. If your exposure choice leads to clipped highlights at base ISO, you need to back off the exposure.
So the choice of ISO for a given exposure is a tradeoff. Let's look at a few cameras (horizontal axis is ISO in stops; first data point is ISO 100 in each case):
(http://theory.uchicago.edu/~ejm/pix/20d/posts/dpr/readnoise_vs_iso.jpg)
For the Canon 5D2 and the Nikon-designed CMOS D3, the camera read noise drops until about ISO 1600 or 3200; for these cameras, you get a S/N benefit from raising the ISO, subject to the constraint of not clipping wanted highlights, up to this point; there is no benefit to raising the ISO beyond this point. For the remaining bottom three examples (D7000, Sony Exmor CMOS sensor; Nikon D2x, CCD sensor; Phase One P65+, CCD sensor), the read noise is relatively flat with ISO, and there is little to be gained in the quality of the raw data from raising the ISO much above the base value.
-
Perhaps a better title for this thread would have been “Stalking the Reichmann Hypothesis”? :)
-
I'd like to decouple the discussion from the histogram, which is where 'To The Right' refers to in ETTR. The most important consideration is always collecting the most photons during the exposure
Perhaps a better title for this thread would have been “Stalking the Reichmann Hypothesis”? :)
Michael's main point that one should maximize exposure is correct, but use of the histogram and the term ETTR can be misleading, as Emil points out above and in his reply to the DitigalDog. By now, everyone seems to agree that the number of levels has little significance.
Regards,
Bill
-
As people have been saying (repeatedly) the issue is S/N ratio, not the number of levels. At a given illumination level (number of photons = S), the noise goes as sqrt(S), so the S/N = sqrt(S). You improve S/N by capturing more photons -- raising the exposure. Regardless how many levels there are, two tones are distinguishable (on average) if they are spaced more than the noise. So there are roughly sqrt(S2)-sqrt(S1) distinguishable tones in the exposure range between S1 and S2. If you increase the exposure by a stop, you increase this number of distinguishable tones by a factor sqrt(2)~1.4, even though the number of levels in the raw data has doubled. Similarly, if you took the same picture with a D700 in 12-bit mode and 14-bit mode, the 14-bit capture would have 4x more levels in the raw data, but the same number of distinguishable tones, because at the same exposure the same number of photons is captured. It's the S/N, not the number of levels.
There are corrections to the above analysis due to read noise, that are only important at low exposure (deep, deep in the shadows at low ISO, creeping upward into higher zones as the ISO is raised).
Emil,
Believe it or not I do understand the principle that the signal comprising the image we capture consists of photons, and that increasing the number of photons during an exposure, by increasing exposure time, increases the strength of the signal and consequently improves the ratio of signal to noise.
I'm not arguing that this is not the case. It's just that I fail to see the logic in the statement that it is not the number of levels also. It seems to me that SNR and the 'potential' number of levels that the camera can record as a result of increasing exposure, go hand in hand.
Now, you have given us an example whereby simply increasing the camera's capacity to record a greater number of levels, by switching from 12 bit mode to 14 bit, may have no visible impact on the image. Fair enough! However the key phrase here is visible impact.
The increased number of levels in a 14 bit capture may be in regions of the tonal range where there may already be more levels in the 12 bit capture than the human eye can distinguish.
If we are to look for such increase in the number of levels, in the 14 bit capture, we stand a better chance of seeing them in the shadows where there are relatively few distinguishable levels in the first instance. Any increase at all may be appreciated.
Now, I understand there are claims that 14 bit in-camera processing in DSLRs is over-kill and that it's sometimes difficult to see any IQ benefit at all because maybe 12 bit is already sufficient. But, surely this is a different issue.
Supposing we were able to compare 6 or 8 bit captures with 12 or 14 bit captures in a modern DSLR, using identical exposures. Now I know that no manufacturer would waste their time providing the consumer with a 6 or 8 bit option in a modern DSLR, but my point is, if they were to, I think you would actually be able to discern more levels in the 12 bit capture using the same exposure.
Finally, if you wish to claim that the number of levels is not the issue with ETTR, perhaps you can provide an example of a comparison demonstrating that an increase in exposure, resulting in a better SNR, has not also resulted in an increase in the number of levels. That could be a convincing demonstration that the number of levels is not the issue.
But, before you attempt this, there should be a few ground rules. It is understood that the levels to be distinguished must exist in the scene being captured, across the whole tonal range, and must be greater in number than the eye can distinguish, across that entire tonal range, otherwise the demonstration would be flawed. There would be no point in photographing a large grey card consisting of one level, say 128, 128, 128.
Do I get the prize for obfuscation or for clarity? ;D
-
To add even greater clarity to the issue, perhaps the following could be considered true.
Increasing exposure towards an ETTR, just short of highlight clipping, will always result in a better SNR whatever the nature of the scene. However, the increased number of visually distinguishable levels in the resultant image may not be greater, if such differences in levels do not exist in the first instance, in the scene being photographed.
The primary effect of increasing exposure is therefore a better SNR. A secondary consequence is a potential increase in the number of levels recorded, and the number of levels visually distinguishable.
-
Are the words "levels" and "tones" being used interchangeably ?
Frank
-
Are the words "levels" and "tones" being used interchangeably ?
Frank
I believe so. 128, 128,127 is one level, or tone, and 127,128,127 is another. However, I doubt very much that these would be visually distinguishable, side by side.
-
i think Emil may have a slightly different definition if i understand what he is saying correctly from the quotation above -
" If you increase the exposure by a stop, you increase this number of distinguishable tones by a factor sqrt(2)~1.4, even though the number of levels in the raw data has doubled."
Frank
-
Finally, if you wish to claim that the number of levels is not the issue with ETTR, perhaps you can provide an example of a comparison demonstrating that an increase in exposure, resulting in a better SNR, has not also resulted in an increase in the number of levels. That could be a convincing demonstration that the number of levels is not the issue.
I don't have the equipment, the time, or the interest to go off chasing this particular wild goose, but it is certainly possible to make such a demonstration with currently available gear. For instance, take a D7000; take several images of the same scene:
1) ISO 4X, with exposure Y (in EV), with no clipped highlights.
2) ISO X, with exposure Y (thus 'ETTL' by two stops, using lower ISO).
3) ISO 4X, with exposure Y-2EV (thus 'ETTL' by two stops using two stops less exposure).
(2) and (3) will have the same number of levels for any given patch of the image, but I guarantee (2) will have higher quality because it has higher S/N due to higher exposure. I also guarantee that (1) and (2) will be indistinguishable, even though (2) has 4x fewer levels on any patch of the scene (I might ask that X=200 or higher to avoid issues in the 12th or 13th stop below clipping ;) ). If you want higher exposure with fewer levels, try comparing eg example (1) with ISO X/2, exposure Y+2EV. The latter will have twice the S/N but half the levels in any image patch, compared to (1). Conversions need a level playing field -- the raw converter can't monkey with conversion parameters based on ISO in metadata. RawTherapee 4.0 is acceptable for this purpose.
-
I believe so. 128, 128,127 is one level, or tone, and 127,128,127 is another. However, I doubt very much that these would be visually distinguishable, side by side.
Do the number of 'levels' and 'distinguishable tones' in reply 85 look the same to you?
-
Ray,
Since you are the one out in left field on this one perhaps rather than challenging other people to provide examples to "prove" to you that fundamental image and signal processing concepts are in fact true you could take the time to do them yourself. No one here is served by wasting their time illustrating the same point over and over only to have you plug your ears. If you want to understand something take the time to explore it yourself since clearly no one is going to convince you of anything. If you own a camera and a RAW converter you've already got all the tools you need.
Ken
-
Do the number of 'levels' and 'distinguishable tones' in reply 85 look the same to you?
Of course they don't. There is obviously a distinction to be made between what's recorded and visually distinguishable, and what's recorded but not visually distinguishable.
What I've asked you to provide is a comparison whereby increasing the exposure, and consequently the SNR, does not simultaneously increase the number of distinguishable levels where such greater numbers of different levels do actually exist in the scene being photographed, and also exist in that part of the tonal range where the eye is able to perceive such increase in the number of levels.
If you can do this, then I would concede that ETTR is all about SNR and not about levels.
-
Ray,
Since you are the one out in left field on this one perhaps rather than challenging other people to provide examples to "prove" to you that fundamental image and signal processing concepts are in fact true you could take the time to do them yourself. No one here is served by wasting their time illustrating the same point over and over only to have you plug your ears. If you want to understand something take the time to explore it yourself since clearly no one is going to convince you of anything. If you own a camera and a RAW converter you've already got all the tools you need.
Ken
You're joking, aren't you? I've done hundredes of tests over the years comparing such issues with different exposures, different cameras and different ISOs.
The most recent set of experiments I made was when I took delivery of the D7000, having bought the camera because I was intrigued by its high DR capability. I took dozens of test shots ranging from full ETTR exposures to 14 stops underexposure.
With each stop of increase in exposure, from 14 stops underexposed, I've observed an increase in the number of distinguishable levels of detail, where detail exists in the real scene. Where detail doesn't exist, as in a plain wall, then of course one doesn't expect to see an increase in the number of 'real' tones, merely a reduction in the number of spurious tones, ie, noise.
-
What I've asked you to provide is a comparison whereby increasing the exposure, and consequently the SNR, does not simultaneously increase the number of distinguishable levels where such greater numbers of different levels do actually exist in the scene being photographed, and also exist in that part of the tonal range where the eye is able to perceive such increase in the number of levels.
If you can do this, then I would concede that ETTR is all about SNR and not about levels.
The discussion has been about the assertion in Michael's posts on ETTR that the benefit derives from the increase in available raw levels, not from the increase in distinguishable tonalities resulting from increased S/N. I don't know why you are trying to change the basis of the discussion. Of course increasing the SNR increases the number of distinguishable tonal values; that was my point all along, so of course I won't be providing you a counterexample. The issue was always whether the image quality improvement had anything to do with the number of raw levels used to represent the image, and I have provided a wide variety of counterexamples to that assertion. Since you have a D7000, I suggest you try my proposed experiment in reply #98.
-
The discussion has been about the assertion in Michael's posts on ETTR that the benefit derives from the increase in available raw levels, not from the increase in distinguishable tonalities resulting from increased S/N. I don't know why you are trying to change the basis of the discussion. Of course increasing the SNR increases the number of distinguishable tonal values; that was my point all along, so of course I won't be providing you a counterexample. The issue was always whether the image quality improvement had anything to do with the number of raw levels used to represent the image, and I have provided a wide variety of counterexamples to that assertion. Since you have a D7000, I suggest you try my proposed experiment in reply #98.
Emil,
The argument is really about semantics. I'm not in any fundamental disagreement with you. I just see things from a slightly different perspective, perhaps.
The initial effect of increasing exposure is an improvement in SNR. There no doubt about it.
The consequences of this increase in SNR is that more detail (or a greater number of tones or levels) become apparent, no longer obscured by noise.
From my very pragmatic perspective, this increase in detail is what's important because I know if it's lost, or buried too deep in the noise, it can't be retrieved.
However, if there's noise present where no detail exists, then any good noise reduction program can fix the problem. One can have a severely underexposed shot of a lady's face and make it appear as smooth as a baby's bottom, with a good noise reduction program, but at the expense maybe of losing a few eyelashes.
It's the increase in levels, where it can be seen, which is the main benefit of ETTR for us pragmatic photographers.
-
You're joking, aren't you?
No I really wasn't.
I've done hundredes of tests over the years comparing such issues with different exposures, different cameras and different ISOs.
Then you shouldn't have any trouble doing a few more.
The most recent set of experiments I made was when I took delivery of the D7000, having bought the camera because I was intrigued by its high DR capability. I took dozens of test shots ranging from full ETTR exposures to 14 stops underexposure.
Awesome, you should already have your answer then. If you want more, you've got the best camera around to perform the tests Emil suggested.
With each stop of increase in exposure, from 14 stops underexposed, I've observed an increase in the number of distinguishable levels of detail, where detail exists in the real scene. Where detail doesn't exist, as in a plain wall, then of course one doesn't expect to see an increase in the number of 'real' tones, merely a reduction in the number of spurious tones, ie, noise.
Yep, that jives with what everyone is saying and what you've been saying as well as far as SNR goes.
The final step to your question about RAW levels is a trivial one with your D7000. Take one exposure at 0 eV. Take two exposures at -1 EV, average the two RAW files, do +1 EV compensation on the averaged RAW file. Take four exposures at -2 EV, average the four RAW files, do +2 EV on the averaged RAW file. Are there any differences in the mid tones of these three scenarios? If yes, RAW levels matter, if no RAW levels don't matter only SNR. If you can't average RAW files on your own the application PhotoAcute will do it for you and has a free trial download with a bit of water marking you can work around for this test.
Let us know what you find out.
Ken
P.S. Apologies if I've been missing what you are asking in all the semantics - the thread is confusing to say the least.
-
The final step to your question about RAW levels is a trivial one with your D7000. Take one exposure at 0 eV. Take two exposures at -1 EV, average the two RAW files, do +1 EV compensation on the averaged RAW file. Take four exposures at -2 EV, average the four RAW files, do +2 EV on the averaged RAW file. Are there any differences in the mid tones of these three scenarios? If yes, RAW levels matter, if no RAW levels don't matter only SNR. If you can't average RAW files on your own the application PhotoAcute will do it for you and has a free trial download with a bit of water marking you can work around for this test.
Let us know what you find out.
Why would I want to do that? There are all sorts of procedures one could adopt in order to get a particular result. I'm only concerned with what's of benefit in the field or the studio. I'm not a laboratory technician.
I once spent some time testing the stacking feature in Photoshop Extended. The principle is, one takes a number of shots of a static scene, say 5 or 8. When the images are stacked using the appropriate mode in Photoshop, the 'best' pixels from each frame are automatically selected to result in an image with lower noise and better detail. The improvement is very obvious at high ISO, with a reduction in noise and a corresponding increase in detail (number of 'real image' levels) equivalent to about 2 stops, maybe a bit less.
The issue for me is, in what circumstances would this technique be useful in the field. I could find only one circumstance where this might be useful, and that was when the scene was static, I had no tripod, and the light was poor, requiring use of a high ISO for a reasonably fast shutter speed to freeze camera movement.
If I were really careful, with steady hand, perhaps leaning against a tree, I could take about 5 shots at say ISO 1600 and 1/25th, and get a 'stacked' result roughly equivalent to a single ISO 400 shot taken on a tripod at a shutter speed of 1/6th.
P.S. Apologies if I've been missing what you are asking in all the semantics - the thread is confusing to say the least.
Apologies accepted, but why are you confused? Noise is a problem because it obscures detail (read levels). If there's no detail to be obscured, then ETTR is surely less of a concern because any unwanted, visible noise can be easily removed in noise reduction programs, or even the noise reduction slider in ACR.
-
If yes, RAW levels matter, if no RAW levels don't matter only SNR.
We already know the answer to that, based on literally hundreds of thousands of exposures by some of the most picky IQ conscious photographers in the world. Shoot a Leica M9 in compressed mode, and your mid range f/stop (zone 5) has 14 raw levels. Yes, that's not a mistake, 14 levels. Gamma encoded, sure, so those levels are optimally distributed in the zone, but still only 14 discrete raw levels. Uncompressed you have 256 raw levels in the same zone. Nobody that I know of has ever been able to demonstrate that in a practical shooting situation the M9's uncompressed image has visibly better IQ than the compressed image.
I don't know whether Emil's numbers for the P65 are right or not, but they demonstrate a basic point - if you can only distinguish 10 tones in a zone, then whether you encode those 10 tones into 20 raw levels or 1000 raw levels make no difference to visible IQ at all.
It's probably possible to construct scenarios where raw levels matter, if you tried hard enough. But in a real world situation shooting a modern camera with a real sensor with real noise? Raw levels don't matter.
Sandy
-
The issue for me is, in what circumstances would this technique be useful in the field. I could find only one circumstance where this might be useful, and that was when the scene was static, I had no tripod, and the light was poor, requiring use of a high ISO for a reasonably fast shutter speed to freeze camera movement.
Shooting a person/animal in low light using a stand, to improve the IQ of the background?
-h
-
It's the increase in levels, where it can be seen, which is the main benefit of ETTR for us pragmatic photographers.
Ray,
You are truly a bitter-ender ;). How are you quantifying all those levels that are overlapped by noise and how do you know that the improvement in IQ is due to increased levels and not merely due to improved SNR?
Bill
-
It's the increase in levels, where it can be seen, which is the main benefit of ETTR for us pragmatic photographers.
I have been hesitating about two possible answers to this, so I'll answer twice:
ANSWER #1
Have you ever _seen_ any increase in levels thanks to ETTR?
ANSWER #2
(http://www.guillermoluijk.com/misc/maemia.jpg)
-
Ray,
You are truly a bitter-ender ;). How are you quantifying all those levels that are overlapped by noise and how do you know that the improvement in IQ is due to increased levels and not merely due to improved SNR?
Bill
Ah! Good question, Bill. I think I've already agreed that the improvement in IQ, when increasing exposure, is a result of an improvement in SNR, but such improvement is expressed (or viewed to be relevant and useful) as a discernment in the number of different 'real' tones, or levels in the converted image.
In the absence of sufficiently different levels (detail and texture) in the scene being photographed, one would expect to notice a reduction in levels or tones, that is, a reduction in spurious levels or tones, described as noise, unless one is viewing detail in the upper tonal region where noise may not be a problem and the number of levels recorded is far greater than the eye can discern in any case.
I've raised the question, is it possible to have an image with reduced 'visible' noise, as a result of a greater exposure, which does not allow for the 'visual' discernment of a greater number of levels or tones, when such tones exist in the scene being photographed.
If the answer is 'no', then I think we would have to agree that ETTR is all about improvement in SNR and increased number of levels. No? ;D
-
What I've asked you to provide is a comparison whereby increasing the exposure, and consequently the SNR, does not simultaneously increase the number of distinguishable levels where such greater numbers of different levels do actually exist in the scene being photographed, and also exist in that part of the tonal range where the eye is able to perceive such increase in the number of levels.
I don’t think Michael ever implied that ETTR provides more levels right? I think the point made originally was there are fewer levels in the shadow area of the (linearly) captured data and that by implementing ETTR, you end up with the cleanest (less noise) levels where so few levels originate.
The initial effect of increasing exposure is an improvement in SNR. There no doubt about it.
The consequences of this increase in SNR is that more detail (or a greater number of tones or levels) become apparent, no longer obscured by noise.
I think all of us are in agreement in those comments right? Noise has a tone/level, its non image data (noise, gunk, crap) and doesn’t provide anything useful in terms of showing us what we hoped to capture.
Did we get sidetracked with semantics with the idea that ETTR provides more levels?
-
I have been hesitating about two possible answers to this, so I'll answer twice:
ANSWER #1
Have you ever _seen_ any increase in levels thanks to ETTR?
Ha! ;D All the time, whenever I increase exposure. It's wonderful. But I'm referring to differences in real tones and levels that exist in the scene. The increase in such levels is of course most apparent in the shadows, as you know.
-
I don’t think Michael ever implied that ETTR provides more levels right? I think the point made originally was there are fewer levels in the shadow area of the (linearly) captured data and that by implementing ETTR, you end up with the cleanest (less noise) levels where so few levels originate.
I think all of us are in agreement in those comments right? Noise has a tone/level, its non image data (noise, gunk, crap) and doesn’t provide anything useful in terms of showing us what we hoped to capture.
Did we get sidetracked with semantics with the idea that ETTR provides more levels?
Well, he starts out with the passage:
A typical consumer DSLR recording 12 bits per sensel is able to record up to 4,098 separate tonal values.
If we assume a 10 stop dynamic range this is how this data is distributed...
The brightest stop = 2048 tonal values
The next brightest stop = 1024 tonal values
The next brightest stop = 512 tonal values
The next brightest stop = 256 tonal values
The next brightest stop = 128 tonal values
The next brightest stop = 64 tonal values
The darkest stop = 32 tonal values
As can be seen, each stop from the brightest to the darkest contains half of the data of the one preceding it.
This helps explain why noise is seen most in the darkest areas of a file. In the brightest areas there is a lot of data and so the noise floor (which is always present) only represents a small percentage of the total signal (or data). In the darker areas, where data is sparse, ever-present noise becomes easily visible.
This implies that noise has something do to with the number of levels. Later in the essay the does a SNR analysis using photon shot noise as a limiting factor. However, he never quite comes around to the fact that the number of discrete and distinguishable levels is limited by noise throughout the image from highlights to shadows (disregaring read noise) rather than the number of levels in the raw file. Much confusion would have been eliminated had he left out that passage and gone directly to SNR analysis.
Regards,
Bill
-
Did we get sidetracked with semantics with the idea that ETTR provides more levels?
I think so. I'm making a distinction between the recording of real levels in the scene, as opposed to spurious levels defined as noise.
I have a few images which are drastically underexposed, yet contain more levels than some ETTR exposures. The levels take the form of pattern noise which looks like a coarse jute fabric, quite detailed.
See attached image. With a greater exposure, that sky would have lost a great number of different levels. ;D
-
Did we get sidetracked with semantics with the idea that ETTR provides more levels?
Actually, it's not a sidetrack, it's central to the ETTR argument - without the alleged benefits from the additional raw levels, ETTR is a waste of time outside of certain special situations (synthesizing a lower ISO, optimizing read noise versus sensor noise for some cameras). But the general case for ETTR just vanishes without levels because you can get the S/N advantages of ETTR by just changing ISO, and exposing normally.
Sandy
-
I don’t think Michael ever implied that ETTR provides more levels right? I think the point made originally was there are fewer levels in the shadow area of the (linearly) captured data and that by implementing ETTR, you end up with the cleanest (less noise) levels where so few levels originate.
Did we get sidetracked with semantics with the idea that ETTR provides more levels?
The quibble is that in both posts, Michael puts the following statement -- which is misleading at best -- front and center:
A typical consumer DSLR recording 12 bits per sensel is able to record up to 4,098 separate tonal values.
* If we assume a 10 stop dynamic range this is how this data is distributed...
o The brightest stop = 2048 tonal values
o The next brightest stop = 1024 tonal values
o The next brightest stop = 512 tonal values
o The next brightest stop = 256 tonal values
o The next brightest stop = 128 tonal values
o The next brightest stop = 64 tonal values
o The darkest stop = 32 tonal values
* As can be seen, each stop from the brightest to the darkest contains half of the data of the one preceding it.
and a little bit later:
The reason why we want to expose every shot that we take with the data as far to the right of the histogram as possible is because that's where the data is! It also is where the visible noise isn't. The visible noise is lurking in the darker stops.
Since both increasing the exposure and increasing the ISO move the histogram to the right, one is led to conclude that they are equally good according to this criterion, and this is simply false, as one sees from the correct analysis in terms of S/N. The conflating of S/N and number of raw levels is all over Michael's writing on this subject, and it has done a tremendous disservice to the photography community. For instance, for many cameras (including all MFDB's; I showed which ones in post #90), ETTR doesn't make sense above base ISO -- you are better off at or near base ISO* even if the maximum exposure you can tolerate leaves the histogram to the left. With these cameras, the data has the same shadow/midtone quality at the lower ISO, with the histogram to the left, and more headroom for accurately capturing highlights, relative to the same maximum tolerable exposure at a higher ISO where the histogram would be more to the right.
*Since MFDB's typically use 16-bit data output from a system with 12-bit DR, the data is already overspecified and so base ISO is ample.
Consequently, the whole section of Michael's most recent post entitled "Welcome to the 21st Century" is misguided for the above class of cameras. For such cameras, the proper optimization is not an 'ETTR mode' which sets the exposure on the basis of (a user specified) percentage of clipped pixels. Rather, one wants a 'maximize exposure' mode. How would this work for this class of cameras? It would act much like "safety shift" does on Canon DSLRs -- it sets the ISO to base; the user specifies their max tolerated exposure by setting Tv and Av according to their motion blur and DoF requirements (ie as in Manual mode); and if a specified threshold percentage of pixels is clipped, the exposure is reduced. This functionality is rather different from the proposals in Michael's post, unless the exposure set is such that the safety shift is always invoked.
-
Well, he starts out with the passage:
The darkest stop = 32 tonal values
As can be seen, each stop from the brightest to the darkest contains half of the data of the one preceding it.
This helps explain why noise is seen most in the darkest areas of a file. In the brightest areas there is a lot of data and so the noise floor (which is always present) only represents a small percentage of the total signal (or data). In the darker areas, where data is sparse, ever-present noise becomes easily visible.[/i]
This implies that noise has something do to with the number of levels.
The key word there is implies. He doesn’t say as far as I can see, that ETTR alters the numbers of levels, only provides cleaner (less noise) in those levels.
-
Shooting a person/animal in low light using a stand, to improve the IQ of the background?
-h
How does that work? If one takes half a dozen shots over a period of one second, the person or animal will likely have moved. The person or animal will be blurred.
-
The key word there is implies. He doesn’t say as far as I can see, that ETTR alters the numbers of levels, only provides cleaner (less noise) in those levels.
I halfway agree. I never had a hint of implication from Michael's text that the number of levels were altered, or changed in any way...nothing seemed to be implied at all. The understanding that I received was very clear and logical to me. I immediately put my camera on a tripod and started shooting and testing the idea. I was very impressed and felt like I had learned an extremely important bit of knowledge that would improve my chances of making the best capture at any given moment.
I think that a lot of very sharp minds have been wasting a lot of time over-analyzing Michael's statements. Instead, they should have been out exploring exactly how to incorporate and implement the information into their own workflow and photography life.
-
I think that a lot of very sharp minds have been wasting a lot of time over-analyzing Michael's statements. Instead, they should have been out exploring exactly how to incorporate and implement the information into their own workflow and photography life.
Yep...
-
I think that a lot of very sharp minds have been wasting a lot of time over-analyzing Michael's statements.
Depends whether you think misinformation is helpful in optimizing your technique.
Instead, they should have been out exploring exactly how to incorporate and implement the information into their own workflow and photography life.
I know exactly what the noise characteristics of my cameras are, how to optimize capture and workflow for best S/N and DR based on that information. Do you? For me the answer is more subtle than "ETTR", and depends inherently on those measured noise characteristics.
-
I know exactly what the noise characteristics of my cameras are, how to optimize capture and workflow for best S/N and DR based on that information.
I must say that phrases like the above make me want to run away and hide. I don't have any interest whatsoever in optimising capture and workflow, thanks very much. I want to take good photographs, and print better pictures. Now you might say that the two things are the same, and that I am quibbling about semantics, but I don't think so. I see photography as an essentially artisitic activity, and the language we use to describe the process should not alienate others in a babble of techno-jargon.
John
-
Depends whether you think misinformation is helpful in optimizing your technique.
Sometimes, misinformation can help me to optimize my technique. This is due to the fact that in proving the information to wrong (for myself) I can solidify sound techniques for myself. However, I do not see, in any way, how Michael's original article put forth misinformation. I still think that a lot of "misanalyzing" has occurred.
I know exactly what the noise characteristics of my cameras are, how to optimize capture and workflow for best S/N and DR based on that information. Do you? For me the answer is more subtle than "ETTR", and depends inherently on those measured noise characteristics.
I am learning the noise characteristics of my current camera, just as I have been doing with each body since my first Kodak in 1998. I am still learning how to make my captures and workflow better in regards to Signal to Noise Ratio and Dynamic Range based on the knowledge that I have gained from experience as well as from reading the findings of others. So no, I have not reached the level of pompous perfection in my mind to declare that I have mastered it and know exactly all that there is to know. For me, the answer may end up being more subtle than "ETTR", but for now, I am happy to have the information in my noggin' and to have the enjoyment of using it. This is truly a learning forum, and sometimes, I learn things that are completely untechnical and unphotographic.
-
In my opinion, for some of us, image making is about the science. For others it is about the art. For most of us it is somewhere between and we can learn from both. While we may take offense at some writings of others, there is no reason for incivility. It only gets in the way of learning and discourages participation.
Thanks.
Frank
-
In my opinion, for some of us, image making is about the science. For others it is about the art. For most of us it is somewhere between and we can learn from both. While we may take offense at some writings of others, there is no reason for incivility. It only gets in the way of learning and discourages participation.
Thanks.
Frank
Frank, you are absolutely right. Mine was a poorly judged post, and I apologise for the incivility.
John
-
I think perhaps a current article in Scientific American, titled "Lessons from Sherlock Holmes", is relevant to this current debate.
Here's what Sherlock Holmes has to say in a "Study in Scarlet".
“I consider that a man’s brain originally is like a little empty attic, and you have to stock it with such furniture as you choose. A fool takes in all the lumber of every sort that he comes across, so that the knowledge which might be useful to him gets crowded out, or at best is jumbled up with a lot of other things, so that he has difficulty laying his hands upon it.
Now the skillful workman is very careful indeed as to what he takes into his brain-attic. He will have nothing but the tools which may help him in doing his work, but of these he has a large assortment, and all in the most perfect order. It is a mistake to think that that little room has elastic walls and can distend to any extent. Depend upon it there comes a time when for every addition of knowledge you forget something that you knew before. It is of the highest importance, therefore, not to have useless facts elbowing out the useful ones.”
http://blogs.scientificamerican.com/guest-blog/2011/08/26/lessons-from-sherlock-holmes-cultivate-what-you-know-to-optimize-how-you-decide/
-
Depends whether you think misinformation is helpful in optimizing your technique.
I know exactly what the noise characteristics of my cameras are, how to optimize capture and workflow for best S/N and DR based on that information. Do you? For me the answer is more subtle than "ETTR", and depends inherently on those measured noise characteristics.
BUT ..... does that make you a good photographer or someone who is knowledgeable about the process involved? ;)
-
Frank, you are absolutely right. Mine was a poorly judged post, and I apologise for the incivility.
John
Why the apology John? You said what a lot of the members are probably thinking, unless of course you are being facetious. :-\
-
In my opinion, for some of us, image making is about the science. For others it is about the art. For most of us it is somewhere between and we can learn from both. While we may take offense at some writings of others, there is no reason for incivility. It only gets in the way of learning and discourages participation.
Thanks.
Frank
I would have thought it is about the final output? The print or an image posted on the web? I sometimes wonder about the final output of the members who seem obsessed by the science. I don't notice many of them posting their efforts on the critique forum. People like John post fine images but doesn't get caught up in the scientific side of things. He is what I call a practical photographer who knows enough to produce good work. I hope I am not being uncivil? ;)
-
I would have thought it is about the final output? The print or an image posted on the web? I sometimes wonder about the final output of the members who seem obsessed by the science. I don't notice many of them posting their efforts on the critique forum. People like John post fine images but doesn't get caught up in the scientific side of things. He is what I call a practical photographer who knows enough to produce good work. I hope I am not being uncivil? ;)
Reading your reply made me think about a class I took under Art Rainville http://www.studiorainville.com/ (http://www.studiorainville.com/) . He was teamed up with another photographer who was real technical, used all of the latest high tech equipment (made wonderful captures too) and really preached being exact with everything. So, the first four hours, we heard all of this really good technical info and then comes Art. He moves his lights around, sets the power on the back of the monolights, sticks his index finger into his mouth to wet it, holds it up in front of his subject like he is checking for wind direction, and trips the lights! He never used a light meter during the next four hours. He used his experience and his camera's histogram.
I think that in the end, we all (including myself) need to accept that there is not a single "best" way for everyone to approach anything photographic. There is, however, a single best way for each individual to approach photography: Keep an open mind in order to absorb as much information as possible while learning what specific information is of the best use in your workflow. Being iron-fisted and declaring that only one way exists is not conducive to anyone learning....included the one with the iron fist. A closed mind will learn nothing.
-
BUT ..... does that make you a good photographer or someone who is knowledgeable about the process involved? ;)
In fact he is a very knowledgeable person.
I sometimes wonder about the final output of the members who seem obsessed by the science. I don't notice many of them posting their efforts on the critique forum. People like John post fine images but doesn't get caught up in the scientific side of things. He is what I call a practical photographer who knows enough to produce good work. I hope I am not being uncivil? ;)
Not incivil, just not very clever since science can be a goal on its own. I'm pretty sure most engineers working for camera makers produce bad outputs or even don't take pictures at all, but they are involved in contributing to photography science and hence to photography in general. This thread is about a tecnique to scientifically improve taking pictures, so there is room for both sides of the story.
-
One point not mentioned. The thrust of the argument is about noise in shadows and how to lighten them? What happens if you are like me and like shadows that are black - or nearly black - and a little clipped. I suspect there are a lot of photographers like me. A few years ago the info in Photoshop books was to set blacks at 15 15 15 and whites at 245 245 245. You see very few advocating that now.
-
In fact he is a very knowledgeable person.
Not incivil, just not very clever since science can be a goal on its own. I'm pretty sure most engineers working for camera makers produce bad outputs or even don't take pictures at all, but they are involved in contributing to photography science and hence to photography in general. This thread is about a tecnique to scientifically improve taking pictures, so there is room for both sides of the story.
At the end of the day this is primarily a photographic site. With regards to the scientific discussions I see very little consensus among the advocates and the photographer who has a more practical mind must be a little dismayed at the outcome of some of the threads. However I recognize that some of it is enlightening and I will still read them. Splitting the wheat from the chaff is difficult. :)
-
One point not mentioned. The thrust of the argument is about noise in shadows and how to lighten them? What happens if you are like me and like shadows that are black - or nearly black - and a little clipped. I suspect there are a lot of photographers like me. A few years ago the info in Photoshop books was to set blacks at 15 15 15 and whites at 245 245 245. You see very few advocating that now.
Clipping shadows is a matter of post processing, not of capture. Unlike the highligths, in the RAW file shadows are not clipped, they just have more or less noise according to exposure. So you cannot take any decision about clipping the shadows at shooting time (when ETTR is applied), that will be done later.
The only reason I can see for that (15,15,15) and (245,245,245) recommendation is to prevent whites and blacks loose details in the output device (typ. prints). With properly calibrated equipment and a correct workflow, I see no reason for those safety ranges.
Splitting the wheat from the chaff is difficult. :)
Not more difficult than splitting the wheat from the chaff when looking at the outputs produced by practical mind photographers. In my case, I don't think even 10% of the pictures I see are of any interest to me.
-
In my case, I don't think even 10% of the pictures I see are of any interest to me.
Are you speaking of only your own pictures, only those of others, or both? Just curious.
-
Are you speaking of only your own pictures, only those of others, or both? Just curious.
Of both.
-
The only reason I can see for that (15,15,15) and (245,245,245) recommendation is to prevent whites and blacks loose details in the output device (typ. prints). With properly calibrated equipment and a correct workflow, I see no reason for those safety ranges.
Exactly. The reason we don’t see people recommending arbitrary values is because no set of values is ideal for all images and processes.
But some love the idea of distilling complex items like this into “7 simple points” and dumbing down every process. Kind of reminds me of the old saying about making something fool proof.
-
At the end of the day this is primarily a photographic site. With regards to the scientific discussions I see very little consensus among the advocates and the photographer who has a more practical mind must be a little dismayed at the outcome of some of the threads. However I recognize that some of it is enlightening and I will still read them. Splitting the wheat from the chaff is difficult. :)
Since this is primarily a photographic site, discussions about the science and technology underlying image-making are entirely apposite. Understanding the physics of vibrating strings may not make you a better musician but that understanding is valuable to those designing/refining the instruments. Furthermore, knowledge of the underlying science can "only add. I don't understand how it subtracts." That quote from -
http://www.youtube.com/watch?v=hIZhgLKSBaY
-
BUT ..... does that make you a good photographer or someone who is knowledgeable about the process involved? ;)
This is a fallacious argument, unless you mean to suggest that knowledge about the process somehow makes you a worse photographer... which I would suggest is a rather silly argument.
-
I've seen a lot of references about "good photographer" or "better photographer" in these ETTR threads.
What is a good photographer? How does one define it? ;D
-
I've seen a lot of references about "good photographer" or "better photographer" in these ETTR threads.
What is a good photographer? How does one define it? ;D
One might say there are as many definitions of a good photographer as there are people in the world who are interested in such matters.
On the other hand, many of those definitions will be essentially identical and indistinguishable from each other, so a more pertinent question might be, how many levels of distinguishable opinion can be expressed and recorded in a particular language. ;D
-
Since this is primarily a photographic site, discussions about the science and technology underlying image-making are entirely apposite. Understanding the physics of vibrating strings may not make you a better musician but that understanding is valuable to those designing/refining the instruments. Furthermore, knowledge of the underlying science can "only add. I don't understand how it subtracts.
It subtracts when people get obsessed with the technicalities and forget about the output, print or web uploads of their images. :)
-
This is a fallacious argument, unless you mean to suggest that knowledge about the process somehow makes you a worse photographer... which I would suggest is a rather silly argument.
There are a lot of people who sound knowledgeable but once you listen to them you find out they are merely repeating what they have read and they don't understand what they have read. Just go through some of the threads on here and you will read statements and contradictions that prove my point. If everyone knew what they were talking about then there would be very few threads on here.
-
I've seen a lot of references about "good photographer" or "better photographer" in these ETTR threads.
What is a good photographer? How does one define it? ;D
Quite easily. You look at their output and if there is a consensus of opinion that says the output is good then they are "good" photographer. Again this is subjective. If it is all in their head and the output is deemed "poor" then they aren't a "good" photographer. Actions speak louder than words. :)
-
It subtracts when people get obsessed with the technicalities and forget about the output, print or web uploads of their images. :)
Yes, it also subtracts when someone gets obsessed with the subjective ouput in a thread that is about the objective input.
There are a lot of people who sound knowledgeable but once you listen to them you find out they are merely repeating what they have read and they don't understand what they have read.
Yes, in fact I see no reason for you to be a perfect candidate.
-
It would be interesting if you could show the members some of your photographic output. I can't remember seeing any of it but I apologise if you have posted some and I have missed it. :) If it was only the graphs then they are very good. If you are so knowledgeable about your photographic theory then why are there so many posts disagreeing with what you state? They all can't be wrong. ;)
-
One might say there are as many definitions of a good photographer as there are people in the world who are interested in such matters.
On the other hand, many of those definitions will be essentially identical and indistinguishable from each other, so a more pertinent question might be, how many levels of distinguishable opinion can be expressed and recorded in a particular language. ;D
Ah, but often the number of levels is obscured by noise, Ray. ;)
Actually, I find the only useful definition of "good photographer" is either "me," or else perhaps "anyone who likes the same subject matter as me and who uses similar techniques to capture it." ;D
-
Quite easily. You look at their output and if there is a consensus of opinion that says the output is good then they are "good" photographer. Again this is subjective. If it is all in their head and the output is deemed "poor" then they aren't a "good" photographer. Actions speak louder than words. :)
Granted, there is more to great photography than knowing how best to use your equipment.
However there is nothing wrong with knowing how to maximize the quality of the information gathered in your image so that you have a better basis to start your post processing.
There is a story, attributed to Bill Fortney, where Bill was asked how he got a great iconic shot of a Galcier National Park sunrise 9or sunset). Bill's answer was....you are not going to like my answer. I have gone to Glacier 22 separate times. That's what it takes to get a great photograph.
John
-
... What is a good photographer? How does one define it?...
Ahhh, yes... one of the great dilemmas of mankind! Some of the brightest tried to answer it:
-
... why are there so many posts disagreeing with what you state? They all can't be wrong...
Stamper, this is a textbook example of a fallacy, a non-sequitur. Rebuttals might start on a high note with Galileo, and end on admittedly low note with this graffiti: "Eat sh*t... thousands flies can't be wrong" ;)
-
This thread was getting rather boring, IMHO. So thank you, Slobodan, for livening it up. And a thanks also to Stamper for goading Slobodan into greater inspiration.
Back on the original topic, I was wondering whether frequent Exposing Too Far to the Right means you are a member of the Tea Party?
-
It may mean you are a pervert or a streaker.
Duhhhhhhhhhh...Sorry Eric, just got it ;D ;D I'm a little slow today, seems to get worse with age.
Rich
-
It would be interesting if you could show the members some of your photographic output.
Interesting for you, I guess. Unfortunately my interest in showing you my photographic output is the same I have in knowing yours, none. I simply try to let you know this thread is not about anyone's output, but about ETTR, a technique to objectively improve quality of RAW capture.
-
Interesting for you, I guess. Unfortunately my interest in showing you my photographic output is the same I have in knowing yours, none. I simply try to let you know this thread is not about anyone's output, but about ETTR, a technique to objectively improve quality of RAW capture.
To achieve better output, in other words a good print. None of it matters unless the final outcome is good? :)
-
Let's have a look at Bill Janes rebuttal of some points in Michael's essay, in reply #114, to see where there might be a flaw in Bill's logic, or an incorrect statement.
In particular, the following comment from Bill looks a bit suspicious:
However, he (Michael) never quite comes around to the fact that the number of discrete and distinguishable levels is limited by noise throughout the image from highlights to shadows (disregaring read noise) rather than the number of levels in the raw file.
What does Bill mean by distinguishable? I presume he means visually distinguishable otherwise there would be no need to write 'discrete and distinguishable'
Now this statement from Bill is contrary to my understanding of the situation, which is that the number of levels the camera can record in the brighter stops, despite the presence of noise, is far greater than the eye can distinguish. However, the number of levels the camera can record in the darker stops is far fewer than the capacity of the eye to detect, even without the presence of noise.
Those who insist that it's only the reduction of noise that counts, not the increase in available levels, should consider what would happen in the situation of a theoretical noise-free camera. Let's imagine that all internal camera noise has been eliminated with sophisticated noise-cancellation techniques, and let's imagine that Emil Martinec has invented a device that fits over the lens to remove all photonic shot noise.
A principle of a 12 bit DSLR is that the brightest stop can contain 2048 recorded levels and the darkest stop just 1 level, giving us a theoretical 12 stop DR.
Does anyone really believe that the human eye can distinguish no more than one level in the deepest shadows of a 12 DR scene?
As I understand, the human eye has a tremendous capacity to adjust to changing lighting conditions. When the gaze shifts from the brightest part of a scene, of 12 stops DR that we are about to photograph, to the darkest part of that scene, the pupil dilates, if not instantaneously then within a few seconds, and we see more detail in that darkest shadow, if the detail exists. Certainly far more than the 1 or 2 or 4 levels that a 12 bit DSLR is able to record in those 3 darkest stops, irrespective of noise considerations.
If we want the camera to record say 16 levels in that darkest stop, which the eye has seen in the real scene and which we want to capture, and which may in fact turn out to be only 8 or 10 levels of real information, the rest being noise, we have to overexpose by 4 stops and sacrifice the highlights, or bracket exposure for merging to HDR.
I maintain that ETTR is all about both noise and the number of levels.
-
Does anyone really believe that the human eye can distinguish no more than one level in the deepest shadows of a 12 DR scene?
As I understand, the human eye has a tremendous capacity to adjust to changing lighting conditions. When the gaze shifts from the brightest part of a scene, of 12 stops DR that we are about to photograph, to the darkest part of that scene, the pupil dilates, if not instantaneously then within a few seconds, and we see more detail in that darkest shadow, if the detail exists. Certainly far more than the 1 or 2 or 4 levels that a 12 bit DSLR is able to record in those 3 darkest stops, irrespective of noise considerations.
At a "given state of adaptation" we can distinguish something like 100:1. In-between adaptation we can distinguish something like a gazillion:1. I presume that the presence of bright white objects cause a temporary/spatial "blindness" to really dark details (close to a bright white star or just after driving into a tunnel, it might not really matter if the black stuff is "sort of black" or "very black").
For a real scene it might depend on how large it is/at what distance, and if the bright parts are evenly intermingled with the dark parts, or if it is clearly separated into large "dark" and "bright" regions.
In practice, paper and displays are LDR devices, and will not render anything HDR. If you do extensice tonemapping to squeeze maximum information out of your image aqcuisition, perception does not matter that much because any camera improvement may give you more details to work on.
-
Two tones can be reliably called different only if their difference exceeds the noise. Here is an example for illustration:
(http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/gradient256-composite-alt.gif)
On the top line, tonal jumps are about 8 levels on the 0-255 scale. The next line superposes noise whose typical fluctuation is 12 levels; the boundaries between different tones melt away -- though the 'scene' of tonal steps has discrete jumps, they cannot be distinguished. The next line increases the size of the tonal jumps to about 36, three times the strength of the noise; in the fourth line, one can begin to make out the separate steps in spite of the noise. (Ignore the last two lines; they are more of an illustration of quantization error than anything else.)
-
Post deleted, as Emil had already responded to Ray's contention.
-
I maintain that ETTR is all about both noise and the number of levels.
In case Emil's example was too abstract, try to decide which of these images has better IQ:
(http://www.guillermoluijk.com/article/ettr3/poster.gif)
Do you maintain that more levels is better? or having more levels can be irrelevant in presence of noise?.
This is what happens in the lowest stops of a RAW file, lack of levels is irrelevant since noise makes posterization impossible. So having few levels is not an issue, the issue is having noise, and hence doing ETTR is justified only by SNR maximization.
Think of this: if you could have more levels but the same noise thanks to ETTR, the RAW file wouldn't be any better (like in the kid's image), so ETTR wouldn't be of any interest. The only reason for ETTR is reducing visible noise.
That is what happens in ISOless cameras like the K5 and D7000: ETTR through pushing ISO produces more levels in the shadows, but since SNR remains nearly the same on those cameras, ETTR through ISO has no practical advantages in them. More levels, same IQ, highlights possibly clipped.
Regards
-
Do you maintain that more levels is better? or having more levels can be irrelevant in presence of noise?.
This is what happens in the lowest stops of a RAW file, lack of levels is irrelevant since noise makes posterization impossible. So having few levels is not an issue, the issue is having noise, and hence doing ETTR is justified only by SNR maximization.
Think of this: if you could have more levels but the same noise thanks to ETTR, the RAW file wouldn't be any better (like in the kid's image), so ETTR wouldn't be of any interest. The only reason for ETTR is reducing visible noise.
That is what happens in ISOless cameras like the K5 and D7000: ETTR through ISO produces more levels in the shadows, but since SNR remains the same on those cameras, ETTR through ISO has no practical advantages in them. More levels, same IQ.
Guillermo,
An excellent demonstration. ETTR through ISO produces more raw levels, but the number of distinguishable levels as defined by Emil remains the same, since it is the exposure that determines noise with these cameras. If one replaces the ETTR dogma with maximizing exposure as Emil suggests, maximal exposure can only be achieved at base ISO. As this thread demonstrates, the number of raw levels is often irrelevant. Michael should restate his ETTR rationale, but by now most readers of this thread should know the true rationale for improved image quality is to maximize exposure.
Regards,
Bill
-
Emil,
If it is true that the number of levels that all major DSLRs can potentially record, whether in the deep shadows, the low tones or the midtones, is always significantly greater than the number of tones that can actually be distinguished through the noise, then I would have to concede the point that these theoretical numbers of levels at any particular stop in the tonal range are not relevant.
Looking again at your table for the P65+ in reply #85, there seems to be a massive 'overkill' of the number of potential levels in relation to the number of distinguishable tones that are visible through the noise.
top stop = 166 distinguishable tones, 32768 levels
2nd stop= 117 distinguishable tones, 16384 levels
3rd stop= 82 distinguishable tones, 16536 levels
4th stop= 57 distinguishable tones, 8192 levels
5th stop= 39 distinguishable tones, 4096 levels
6th stop= 26 distinguishable tones, 2048 levels
7th stop= 17 distinguishable tones, 1024 levels
8th stop= 10 distinguishable tones, 512 levels
9th stop= 6 distinguishable tones, 256 levels
10th stop= 3 distinguishable tones, 128 levels
11th stop= 2 distinguishable tones, 64 levels
12th stop= 0.9 distinguishable tones, 32 levels
Does this surplus of levels in the P65+ RAW file serve any useful purpose? For example, if the files were 12 bit instead of 16 bit, we would have just one level in the 12th stop and 2 levels in the 11th stop, equal to the number of distinguishable tones in the 16 bit RAW file.
Would it then be likely, if the file were 12 bit, there would be no distinguishable tones in the 12th stop or the 11th stop, or maybe just one tone in the 11th stop? In other words, does a surplus of potential levels, in any given stop, increase to some degree the number of distinguishable tones that may be discerned through the noise?
-
In case Emil's example was too abstract, try to decide which of these images has better IQ:
(http://www.guillermoluijk.com/article/ettr3/poster.gif)
Do you maintain that more levels is better? or having more levels can be irrelevant in presence of noise?.
This is what happens in the lowest stops of a RAW file, lack of levels is irrelevant since noise makes posterization impossible. So having few levels is not an issue, the issue is having noise, and hence doing ETTR is justified only by SNR maximization.
Think of this: if you could have more levels but the same noise thanks to ETTR, the RAW file wouldn't be any better (like in the kid's image), so ETTR wouldn't be of any interest. The only reason for ETTR is reducing visible noise.
That is what happens in ISOless cameras like the K5 and D7000: ETTR through pushing ISO produces more levels in the shadows, but since SNR remains nearly the same on those cameras, ETTR through ISO has no practical advantages in them. More levels, same IQ, highlights possibly clipped.
Regards
Guillermo,
Those images are much too small to see any difference, and the current monitor I'm using is not particularly good. At that size, one couldn't tell the difference even between a P&S and a P65+.
By the way, I own a D7000 and have many significantly underexposed images taken at base ISO. Raising exposure always seems to increase the number of distinguishable tones in the shadows.
-
Does this surplus of levels in the P65+ RAW file serve any useful purpose?
It allows to quantize the noise more accurately, and thus will lead to less posterization risk. Keep in mind that these Raw levels will undergo a gamma adjustment (=boosting shadow contrast, and compressing highlight contrast) before being displayed.
Cheers,
Bart
-
Here's one of my underexposed D7000 shots, showing the ACR Window before adjustments (except for highlight recovery for some of the bright lights), followed by the processed image.
-
It allows to quantize the noise more accurately, and thus will lead to less posterization risk. Keep in mind that these Raw levels will undergo a gamma adjustment (=boosting shadow contrast, and compressing highlight contrast) before being displayed.
Not sure that this matters much -- stretching or compressing the levels with a gamma adjustment, the noise is stretched/compressed in proportion and its width in proportion to the level spacing remains intact.
There may be some utility to a bit depth of raw that oversamples the data to a degree, as in audio.
-
Guillermo,
Those images are much too small to see any difference, and the current monitor I'm using is not particularly good.
I think that’s the point! Even on the current monitor I’m using which is particularly good.
-
Those images are much too small to see any difference, and the current monitor I'm using is not particularly good.
Those are synthetic images with the only purpose to didactically demonstrate that more levels not always mean anything useful, they come from the same RAW file. Refer to Emil's numbers to convince yourself about a real case for RAW.
It allows to quantize the noise more accurately, and thus will lead to less posterization risk. Keep in mind that these Raw levels will undergo a gamma adjustment (=boosting shadow contrast, and compressing highlight contrast) before being displayed.
What is better, a computer than can perform a task in one milisecond, or a computer that can do it in one microsecond?. For a human being, both are as good, no matter if the second computer was 1000 times faster.
If posterization is not an issue with any arbitrary exposure (and this is the point of the whole story), it's irrelevant that any higher exposure could theoretically provide less posterization risk.
BTW, if your scene contains a plain colour area (like a sky), the more you expose it, the less noise it will have, and hence the more posterization risk when converting to 8-bit JPEG. I have experimented this issue on HDR interiors, where walls absolutely clean of noise, which sounds like a good idea at first, because of RAW overexposure are quite prone to display bands when converted to JPEG. So funily ETTR could lead more easily to posterization in some cases than a lower exposure (posterization in the highlights due to the absence of noise, never due to the lack of levels in the RAW capture).
I already posted an example of this; no one has posted so far a practical example of the opposite, the Myth of ETTR levels!!!.
Regards
-
Here's one of my underexposed D7000 shots, showing the ACR Window before adjustments (except for highlight recovery for some of the bright lights), followed by the processed image.
And what does this mean? It shows that ETTR is not that necessary with a camera having good noise characteristics. With this camera, slight underexposure may be better than attempted ETTR with clipping of highlights.
Regards,
Bill
-
And what does this mean? It shows that ETTR is not that necessary with a camera having good noise characteristics. With this camera, slight underexposure may be better than attempted ETTR with clipping of highlights.
Regards,
Bill
Bill, It's a bit puzzling that DXO Mark attribute almost a 14 stop DR for the D7000, and a full 14 stops DR for the K5, yet in a 14 bit RAW file there would only be one level available for this 14th stop, 2 levles for the 13th stop, and 4 levels for the 12th stop.
The image of a hotel lobby in Thailand posted above, although apparently considerably underexposed, could be considered an ETTR in relation to the specral highlights of the lamps. Therefore, it seems reasonable that a significant part of that particular image will fall within the 14th to 12th stops.
Are 1 to 4 levels sufficient for the D7000 to make the most of these deep shadows, despite the noise?
-
Bill, It's a bit puzzling that DXO Mark attribute almost a 14 stop DR for the D7000, and a full 14 stops DR for the K5, yet in a 14 bit RAW file there would only be one level available for this 14th stop, 2 levles for the 13th stop, and 4 levels for the 12th stop.
It isn't unusual for DxO to be puzzling since their measurements while useful for comparing sensors to a degree use absolute metrics that really aren't photographically relevant.
For them DR means 0dB SNR in an image resized to 8MP. For the K5/D7000 with a 16MP sensor that means the SNR will be 3dB lower once resized (or equivalently that the DR will be 0.5 EV better once resized, you can see this effect on DxO if you click the "Screen" tab instead of the default "Print" tab, the "Screen numbers will be 0.5EV lower). The quantization noise floor, assuming the ADC is almost saturated when the pixel well is full, would be 6.02dB*(14 bits) + 4.77dB = 89dB at the native 16MP resolution. At 8MP resized it would be 92 dB. Converting back to EV we get 92dB/6.02dB = 15.3 EV as the theoretical maximum DR from a perfect 14-bit ADC perfectly loaded on a 16MP imager using DxO's methods.
The problem is that no photographer is going to find 0dB to be acceptable shadows at all. The practical DR is lower than DxO's 14 EV. They had to chose an SNR level for reference so they used 0dB which is a nice engineering number and what sensors designers usually use to specify pixel DR.
That's a lot of math gobbledygook that probably really isn't worth following as far as helping anyone take nice pictures. The take away is that in general it is best to use the DxO measurements as a relative comparison tool between cameras. The absolute numbers and reference levels they use are a bit arbitrary as far as practical photography goes.
Ken
-
It isn't unusual for DxO to be puzzling since their measurements while useful for comparing sensors to a degree use absolute metrics that really aren't photographically relevant.
For them DR means 0dB SNR in an image resized to 8MP. For the K5/D7000 with a 16MP sensor that means the SNR will be 3dB lower once resized (or equivalently that the DR will be 0.5 EV better once resized, you can see this effect on DxO if you click the "Screen" tab instead of the default "Print" tab, the "Screen numbers will be 0.5EV lower). The quantization noise floor, assuming the ADC is almost saturated when the pixel well is full, would be 6.02dB*(14 bits) + 4.77dB = 89dB at the native 16MP resolution. At 8MP resized it would be 92 dB. Converting back to EV we get 92dB/6.02dB = 15.3 EV as the theoretical maximum DR from a perfect 14-bit ADC perfectly loaded on a 16MP imager using DxO's methods.
The problem is that no photographer is going to find 0dB to be acceptable shadows at all. The practical DR is lower than DxO's 14 EV. They had to chose an SNR level for reference so they used 0dB which is a nice engineering number and what sensors designers usually use to specify pixel DR.
That's a lot of math gobbledygook that probably really isn't worth following as far as helping anyone take nice pictures. The take away is that in general it is best to use the DxO measurements as a relative comparison tool between cameras. The absolute numbers and reference levels they use are a bit arbitrary as far as practical photography goes.
Ken
Ken,
It is quite true that a SNR of 0 db is not useful for photography. As you suggest, this is the engineering dynamic range. The photographic range will be less, but that range depends on how much noise you can tolerate in the shadows. The DXO full SNR plot can be used to determine the photographic DR for any given noise floor, as explained by Emil Martinec in this post (http://www.luminous-landscape.com/forum/index.php?topic=42158.0).
Regards,
Bill
-
Those are synthetic images with the only purpose to didactically demonstrate that more levels not always mean anything useful, they come from the same RAW file. Refer to Emil's numbers to convince yourself about a real case for RAW.
I see! So you've deliberately created a set of artificial circumstances to demonstrate a point which may be valid only within that set of artificial circumstances. Is that correct?
-
The practical DR is lower than DxO's 14 EV. They had to chose an SNR level for reference so they used 0dB which is a nice engineering number and what sensors designers usually use to specify pixel DR.
That's a lot of math gobbledygook that probably really isn't worth following as far as helping anyone take nice pictures. The take away is that in general it is best to use the DxO measurements as a relative comparison tool between cameras. The absolute numbers and reference levels they use are a bit arbitrary as far as practical photography goes.
Ken
Really! The two images below, taken with the D7000 shortly after I received it, indicate that the 14th stop contains useful, althought very limited, visible detail. At least you can see that what I've photographed is a Dynamic Range Chart, and make out a few of the largest numbers.
Perhaps my maths is wrong, but I calculate that a range of exposures from 4 seconds to 1/2000th represents 14 stops.
Despite the fact that the target has been artificially created (by Jonathan Wienke), the shots are still 'real-world' shots.
The shot in the 14th stop seems to me to contain too much detail for a 1 level capacity. So maybe my assumed ETTR exposure at 4 seconds is in fact an overexposure.
-
The shot in the 14th stop seems to me to contain too much detail for a 1 level capacity. So maybe my assumed ETTR exposure at 4 seconds is in fact an overexposure.
One level per pixel is many levels when averaged over many pixels. S/N is scale dependent, this is why DxO's figures differ for 'screen' and 'print' for example. So one level is sufficient for a single pixel, it will yield more than one distinguishable level in a downsampled image such as the one you are displaying. (For example group four pixels; the average of a level which can be zero or one for each pixel now ranges over 0,1/4,1/2,3/4,1 so there are five levels, however S/N only doubles and so only two are distinguishable from noise on average; the analysis scales in a straightforward way to coarser scales.)
-
Really! The two images below, taken with the D7000 shortly after I received it, indicate that the 14th stop contains useful, althought very limited, visible detail. At least you can see that what I've photographed is a Dynamic Range Chart, and make out a few of the largest numbers.
And most photographers would choose a black level to make all that noise disappear and lose the detail. That's all I meant by practical, not that you couldn't see anything. In fact, with a large enough structures you should be able to see features well below 0 dB.
Perhaps my maths is wrong, but I calculate that a range of exposures from 4 seconds to 1/2000th represents 14 stops.
Seems right to me.
The shot in the 14th stop seems to me to contain too much detail for a 1 level capacity. So maybe my assumed ETTR exposure at 4 seconds is in fact an overexposure.
I wouldn't be surprised if your exposures are correct. There should still be details this large visible in an image with a bit depth of one. After all, this very small image uses only 1 bit per pixel and I'm sure we can all tell exactly what it is:
(http://upload.wikimedia.org/wikipedia/commons/1/19/Bilevel_1bit_palette_sample_image_-_gimp_dithered.png)
Also worth remembering most every digital audio output device these days uses a 1 bit digital to analog converter. One bit per sample is plenty of information in the right use and context.
Ken
-
And most photographers would choose a black level to make all that noise disappear and lose the detail. That's all I meant by practical, not that you couldn't see anything. In fact, with a large enough structures you should be able to see features well below 0 dB.
Ken
I understand that subjective preferences may determine the choice of black level, but I don't understand how one can see features below 0dB.
Seems right to me.
I wouldn't be surprised if your exposures are correct. There should still be details this large visible in an image with a bit depth of one. After all, this very small image uses only 1 bit per pixel and I'm sure we can all tell exactly what it is:
That's a good point. But my concern here is about the detail provided by a given number of levels. Would that parrot appear more detailed if we raised the number of levels to 4 bit, despite any noise?
-
One level per pixel is many levels when averaged over many pixels.
Makes no sense to me. One is one, and the avarage of one, is one. In a context of one level, any pixel is either black or white. However, in the deep shadows there can be no white, so one level is either black or near black.
Is this not correct?
I'm deducing that the black equlas 0dB and the near-black is the one level.
-
I understand that subjective preferences may determine the choice of black level, but I don't understand how one can see features below 0dB.
You can see features below 0dB if they are large enough. It is similar to SNR improving when you downsize an image. Your visual system is actually surprisingly good at picking out large patterns masked by fine patterned noise. In this case I meant 0 dB in the DxO context - pixel SNR of 0 dB in an 8MP image. You won't see any fine pixel level detail at say -10 dB but if we have big feature, say one taking up nearly a quarter of the image, that will be visible.
This is the whole issue with terms like "0 dB" or "1-bit", they don't really mean anything on their own unless we have some other context (total number of pixels, size of detail of interest, etc. etc.).
That's a good point. But my concern here is about the detail provided by a given number of levels. Would that parrot appear more detailed if we raised the number of levels to 4 bit, despite any noise?
You'd have to leave out the "despite any noise" part to get a useful answer. Noise is the key here. If there was a 4-bit noise free image certainly finer detail would be visible in the 4-bit image than the 1-bit image. If the image was already very noisy the extra bits wouldn't do any good. That is the case with the K5 though - there is already read noise there larger than the quantization noise of the ADC so extra bits won't buy you anything.
Finally, just to add to the confusion, converting from 4-bits to 1-bit we'd add noise before converting to 1-bit to get the best detail in the final 1-bit image! (look up dithering for the details)
Ken
-
It is quite true that a SNR of 0 db is not useful for photography.
Kindly define what exactly is 0 db SNR for a real image as in photography? if I give you an image of my cat how do you define a "noise floor" for that image?
Sincerely,
Joofa
-
You can see features below 0dB if they are large enough. It is similar to SNR improving when you downsize an image. Your visual system is actually surprisingly good at picking out large patterns masked by fine patterned noise. In this case I meant 0 dB in the DxO context - pixel SNR of 0 dB in an 8MP image. You won't see any fine pixel level detail at say -10 dB but if we have big feature, say one taking up nearly a quarter of the image, that will be visible.
...
Finally, just to add to the confusion, converting from 4-bits to 1-bit we'd add noise before converting to 1-bit to get the best detail in the final 1-bit image! (look up dithering for the details)
I believe that dithered "1-bit" prints will tend to have a SNR of ~0dB (or less), while still clearly having practical application.
-h
-
Makes no sense to me. One is one, and the avarage of one, is one. In a context of one level, any pixel is either black or white. However, in the deep shadows there can be no white, so one level is either black or near black.
Is this not correct?
I'm deducing that the black equals 0dB and the near-black is the one level.
The point is that noise acts to dither the levels. Suppose the true signal is some value X between zero and one, over a patch of the image (we are going to ignore natural scene variation for the purpose of answering your question). Suppose the noise is of strength N, for example let N be one level. The noise adds a random number roughly between -N and N to X, so that the pixel wants to record some number between X-N and X+N. Of course, the resulting signal plus noise is digitized so the output is either 0 or 1; if the noise is random (uncorrelated from pixel to pixel), the value of X is reflected in the percentage of 1's vs 0's in the patch -- a fraction X of the pixels will be 1 and the rest 0. If we average the levels over a large enough patch, we recover the original signal, even though each individual pixel only recorded 0 or 1.
This is the basic idea that allows one to trade resolution for noise -- and why downsampled images look less noisy. Note also that while the average is more finely graded than steps of one, that doesn't mean we buy anything by making individual pixels record values more finely spaced than the level of noise, because the individual values are jumping around randomly by an amount between -N and N.
-
I see! So you've deliberately created a set of artificial circumstances to demonstrate a point which may be valid only within that set of artificial circumstances. Is that correct?
No it isn't. I created a set of artificial circumstances to demonstrate 'that more levels not always mean anything useful'. Then added that this is what happens in a RAW file, making more levels thanks to ETTR irrelevant, and I told you to look at Emil's numbers for a real case applied to RAW. Why do you make me repeat you I already said? I find it silly and boring to other users.
Regards
-
Kindly define what exactly is 0 db SNR for a real image as in photography? if I give you an image of my cat how do you define a "noise floor" for that image?
SNR=0dB means the stdDev of Noise equals the level of signal (average value in a patch). To have an idea, you can create a (128,128,128) square patch in PS using some linear profile, and add the amount of gaussian noise needed to make its stdDev=128.
(http://www.guillermoluijk.com/article/digitalp02/ruido_0db_12db.jpg)
-
Here is a very low SNR image, and "below one level" to boot. Taken with a GH-2, DxO gives a DR 11.3 EV and the RAW files are 12-bit.
(http://www.kenandchristine.com/photos/i-DKMrL7C/0/O/i-DKMrL7C.jpg)
(http://www.kenandchristine.com/photos/i-Vx38Z7h/0/O/i-Vx38Z7h.jpg)
So that's 14 EV difference in exposure from a camera DxO says has 11.3 EV of DR in a RAW file with only 12-bits. The 4 second exposure RAW file has the highlights just under the clipping level. In the 1/4000th exposure the difference between the highlights and shadows is 0.25 bits at best. Based on the DxO data the SNR (by DxO's definition) for the 1/4000th exposure is about -15 dB. Nonetheless the largest features are visible, and even the outer portions of the resolution star are still distinguishable.
Ken
-
SNR=0dB means the stdDev of Noise equals the level of signal (average value in a patch). To have an idea, you can create a (128,128,128) square patch in PS using some linear profile, and add the amount of gaussian noise needed to make its stdDev=128.
Unfortunately, real photography is not about patches. Is it? Since this is a photography website I would expect you to apply your theory for an image that I have captured. Can you calculate the SNR of this image:
(http://djjoofa.com/data/images/goofscloseup.jpg)
Sincerely,
Joofa
-
Can you calculate the SNR of this image:
If you can provide a noise-free reference, then yes.
-h
-
If you can provide a noise-free reference, then yes.
-h
Do you take a noise-free reference image with your camera before acquiring a "real" image when you are out for photography? Remember, we must deal with real problems, not manufactured problems.
Sincerely,
Joofa
-
Remember, we must deal with real problems, not manufactured problems.
Then why do you want to determine the noise level from an image that makes that determination more difficult?
Cheers,
Bart
-
Referring to the RAW files of my DR tests for the D7000, I see that the last shot in the series, beginning with a 4 secs exposure, was taken at 1/8000th exposure, which represents the 16th stop.
Surprisingly, there's still some broad detail visible through the noise. At least one can read the heading.
-
Do you take a noise-free reference image with your camera before acquiring a "real" image when you are out for photography? Remember, we must deal with real problems, not manufactured problems.
Sincerely,
Joofa
A lot of the maths and physics allowing you to use your camera in the first place was developed without a camera to experiment on - a lot was developed for totally different applications. Does this mean that you rip the corresponding parts out of your camera?
I dont understand your angle. I have real problems (and real joy) when using my camera. Trying to analyze why it behave like it does (and not always how I would like it to) is a real benefit to me, and I believe that my images are better as a result. SNR is just a tool that measure something, dont measure everything, and let me more quickly understand what other people are talking about. I am fully aware that not all noise looks the same, even at a given average level below the signal. Just like I know that a 200HP car may not be the same as another 200HP car. But I still feel that horsepower is relevant to some applications, just like SNR is relevant to some applications. If you disagree, that is ok.
There is of course the possibility that I "invent" a problem that does not manifest itself in my photography. But I can suggest of equally undesirable flaws that stems from blindly observing "real-world" image results without connecting the observations to a sound physical framework and an understanding of how humans work. One is observer-expectancy: people tend to believe that any "tweak" improves the IQ even when it does nothing at all. Another is that in-camera noise-reduction might cause many images to look better, one might think that this camera is less "noisy" than others. But for some images the flaws of the NR might be very visible. And compared to a seemingly equally noisy camera based on superior sensor tech, one should expect the latter to provide better starting-point for raw images in the long term.
-h
-
A lot of the maths and physics allowing you to use your camera in the first place was developed without a camera to experiment on - a lot was developed for totally different applications. Does this mean that you rip the corresponding parts out of your camera?
I dont understand your angle. I have real problems (and real joy) when using my camera. Trying to analyze why it behave like it does (and not always how I would like it to) is a real benefit to me, and I believe that my images are better as a result. SNR is just a tool that measure something, dont measure everything, and let me more quickly understand what other people are talking about.
-h
I find it rather interesting that there is tons of theoretical stuff like SNR, DR, noise floor, etc in this thread, but no method to calculate the SNR for a real image that I acquired? Seems like our theory has some disconnect when applied to photography. Right?
Sincerely,
Joofa
-
I find it rather interesting that there is tons of theoretical stuff like SNR, DR, noise floor, etc in this thread, but no method to calculate the SNR for a real image that I acquired? Seems like our theory has some disconnect when applied to photography. Right?
Sincerely,
Joofa
No, I just think that you are inventing a problem. I think that you are trying to force this discussion in some unknown direction by claiming irrelevant requirements for any topic in digital photography. We know that SNR affects all images, some more visibly than others. The fact that it cannot easily be estimated from a single image is totally irrelevant. This is an internet discussion forum. It is practically impossible to force other people to stop discussing a topic, and if you try you will hopefully be banned.
-h
-
I find it rather interesting that there is tons of theoretical stuff like SNR, DR, noise floor, etc in this thread, but no method to calculate the SNR for a real image that I acquired? Seems like our theory has some disconnect when applied to photography. Right?
Sincerely,
Joofa
You need a reference point and a single image won't supply that.
-
i once heard the following observation -
"All models are wrong, but some are useful."
SNR may not be possible for a single image, but it has been useful for improving capture devices and thus the images we get from them.
Frank
-
No, I just think that you are inventing a problem. I think that you are trying to force this discussion in some unknown direction by claiming irrelevant requirements for any topic in digital photography. We know that SNR affects all images, some more visibly than others. The fact that it cannot easily be estimated from a single image is totally irrelevant. This is an internet discussion forum. It is practically impossible to force other people to stop discussing a topic, and if you try you will hopefully be banned.
-h
So you are the moral police here now. My original query was not even addressed to you. You jumped in and are now hoping to get me banned. My advise to you is that if problem is too hard for you, sit back and enjoy and ride.
Sincerely,
Joofa
-
Unfortunately, real photography is not about patches. Is it? Since this is a photography website I would expect you to apply your theory for an image that I have captured. Can you calculate the SNR of this image:
(http://djjoofa.com/data/images/goofscloseup.jpg)
Sure I can, the RAW file is needed. Of course SNR will be different according to the different areas (higher on higer exposed areas).
It's a matter of locating each area of the scene:
(http://www.guillermoluijk.com/tutorial/histogrammar/resultgamma.jpg)
in EV zones:
(http://www.guillermoluijk.com/tutorial/histogrammar/zonaslumd.gif)
Each of those EV zones will have a different SNR according to the sensor SNR response (SNR curves can be empirically measured, like DxOMark does).
Regards
-
i once heard the following observation -
"All models are wrong, but some are useful."
SNR may not be possible for a single image, but it has been useful for improving capture devices and thus the images we get from them.
Frank
A basic extension of the usual model will reveal that one can have at least a model of SNR of a single image with some assumptions placed. But the intention was not to go into that model. Rather pointing out the fallacy of statements regarding SNR, and errors in SNR calculation when doing certain operations, say downsampling, as pointed out in this thread and various links provided here.
Sincerely,
Joofa
-
Sure I can, but I need the RAW file, and of course SNR will be different according to the different areas (higher on higer exposed areas).
Guillermo, when we quote noise statistics for a sensor, such as say, X electrons for read noise, etc., do we quote several noise figures, and do each of them apply to a different patch/area?
Joofa
-
errors in SNR calculation when doing certain operations, say downsampling, as pointed out in this thread and various links provided here.
SNR changes when changing resolution, that is not an error, it's an statistical fact.
Once I measured SNR curves for several cameras:
(http://www.guillermoluijk.com/tutorial/noisedr/curvassnr.gif)
But since they have different resolutions, I normalised the results for the resolution of one of them (in this case Canon 5D's 12,7Mpx):
(http://www.guillermoluijk.com/tutorial/noisedr/curvassnrnorm.gif)
-
SNR changes when changing resolution, that is not an error, it's an statistical fact.
Once I measured SNR curves for several cameras:
But since they have different resolutions, I normalised the results for the resolution of one of them (in this case Canon 5D's 12,7Mpx):
Yes, SNR changes, that is agreed. But by how much, that is the issue. That normalization is the issue. Recall, SNR means signal to noise ratio. How can you normalize SNR without incorporating any notions about signal. Right?
Joofa
-
Yes, SNR changes, that is agreed. But by how much, that is the issue. That normalization is the issue. Recall, SNR means signal to noise ratio. How can you normalize SNR without incorporating any notions about signal. Right?
The whole process (measuring SNR curves, normalizing,...) is about incorportating notions about signal, I don't understand your question.
In my case I used a simple criteria, but totally consistent with real world downsizing by assuming noise from different pixels adds in quadrature. See Emil's BIG PIXELS vs. small pixels (http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/noise-p3.html#pixelsize).
Regards
-
The whole process (measuring SNR curves, normalizing,...) is about incorportating notions about signal, I don't understand your question.
In my case I used a simple criteria, but totally consistent with real world downsizing by assuming noise from different pixels adds in quadrature. See Emil's BIG PIXELS vs. small pixels (http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/noise-p3.html#pixelsize).
Regards
Please don't pull an Emil on me ;D. The issue here is the signal changes from the first pixel to the last pixel in the image. So while noise might add in quadrature, but the magnitude of noise being added is dependent on the signal and that is changing in a real world image. Emil has historically shown poor recognition of this problem with his analysis being geared towards a naive assumption of flat patches. But real images are not flat patches. And, few people would appreciate, 10 20, 30, ..., or N number of SNRs figures for a single image coming from each flat patch.
Sincerely,
Joofa
-
But real images are not flat patches. And, few people would appreciate, 10 20, 30, ..., or N number of SNRs figures for a single image coming from each flat patch.
Once the sensor response is obtained, patches are not needed any more. On post 198 I showed a way to relate each area of an image to a SNR in that area.
-
So you are the moral police here now.
In no way, where are you getting that idea from?
My original query was not even addressed to you.
Are you suggesting that any thread should have only two participants?
You jumped in and are now hoping to get me banned.
I think that is a misinterpretation of the post that you are quoting. You seem to forget the part concerning: "if you try to force other people to stop discussing a topic..." What do you think is a fair response if any user tries such a thing?
My advise to you is that if problem is too hard for you, sit back and enjoy and ride.
As far as I have seen, you have just repeated the same question over and over: how to estimate SNR from a single real-world photography. It seems to you have some absolute requirements that all real photographers or participants should adhere to.
...Since this is a photography website I would expect you to...
...Remember, we must deal with...
I have tried to explain why I think it is impossible and fundamentally irrelevant in several different ways.
-h
-
Once the sensor response is obtained, patches are not needed any more. On post 198 I showed a way to relate each area of an image to a SNR in that area.
Yes, you are correct, and that is a way. And, I agree with that. However, as I said I don't know how many people with appreciate 10, 20, 30, ..., SNRs figures attached to a single image. IMHO, this is not a very conducive situation. We should study SNR more closely and figure out models that yield a single number that we can quote for an image with words to the effect of "this image of yours has this SNR", as opposed to "this image of yours has SNRs of 10db, 0db, 3.2 db, 4db, ....., and wait, finally 6db".
Sincerely,
Joofa
-
I have tried to explain why I think it is impossible and fundamentally irrelevant in several different ways.
Then you need to study this problem more.
Sincerely,
Joofa
-
However, as I said I don't know how many people with appreciate 10, 20, 30, ..., SNRs figures attached to a single image.
Not many people are really into photography. Not many photographers choose to spend time on lula. Not many lula-ers contribute to this thread about ETTR... The fact that some topic interest very few people does not make it fundamentally uninteresting, I think.
We should study SNR more closely and figure out models that yield a single number that we can quote for an image with words to the effect of "this image of yours has this SNR", as opposed to "this image of yours has SNRs of 10db, 0db, 3.2 db, 4db, ....., and wait, finally 6db".
If the phenomenon we are studying is complex, it can probably only be described with some complexity. Of course, there may be some simpler phenomenon at its core that is obsured by our observations and understanding, but we wont know until we.. know.
Is the practical application to brag about how much better "SNR" your image has than that of your pal? Or is it to better understand the perceptual gains of doing an optimal exposure? Or something entirely different?
-h
-
Then you need to study this problem more.
I think that this line of communication is not really helping either of us.
-h
-
I said I don't know how many people with appreciate 10, 20, 30, ..., SNRs figures attached to a single image. IMHO, this is not a very conducive situation. We should study SNR more closely and figure out models that yield a single number that we can quote for an image with words to the effect of "this image of yours has this SNR", as opposed to "this image of yours has SNRs of 10db, 0db, 3.2 db, 4db, ....., and wait, finally 6db".
That is possible too: if we can calculate the SNR of each area of the image, it's trivial to calculate the mean, the median of even the histogram of the different SNR found in the image.
But IMO this summarized 'SNR single figure' would only be useful from an engineering point of view. In practice, every photographer processes his pictures in several ways that make that single SNR figure quite theoretical: denoising plain areas will improve perceived SNR, contrast (slopes >1) applied to some area will make SNR worst in that area, while de-contrast (slopes <1) curves will improve it. Darkened areas (no matter if they keep or even improve their SNR), will be less visible to the observer and hence less relevant to the final 'noisy' perception while SNR in the medium and highlights will contribute more to the final perception of noise, etc...
Regards
-
That is possible too: if we can calculate the SNR of each area of the image, it's trivial to calculate the mean, the median of even the histogram of the different SNR found in the image.
True, I agree with you. A weighted average would make sense. But there are fundamental problems regarding patch extraction. Segmentation of images to obtain patches is possible, but perhaps is not always a well defined process, and certain assumptions have to be imposed regarding when pixels are sufficiently different to stop classifying them to the same patch. In any case, if we go this patch-based-segmentation route, do you think it can answer other fundamental questions such as SNR variation in image resolution changes, etc., that you mentioned above?
Sincerely,
Joofa
-
Segmentation of images to obtain patches is possible, but perhaps is not always a well defined process, and certain assumptions have to be imposed regarding when pixels are sufficiently different to stop classifying them to the same patch. In any case, if we go this patch-based-segmentation route, do you think it can answer other fundamental questions such as SNR variation in image resolution changes, etc., that you mentioned above?
To apply the 'patch concept' in measuring SNR we actually don't need real patches, I'd apply a continuos solution instead:
- Calculate exposure on each pixel (from RAW data) with respect to sensor saturation
- Apply gaussian blur (small radius) so that each pixel becomes represented by the average exposure in its surrounding area
- Match those values obtained for each pixel against the SNR response, providing a continuous map of SNR of the RAW file
SNR correction because of resizing is also easy, just needs to use a corrected SNR response. For example to normalise SNR to be comparable to Canon 5D's 12Mpx, I applied this simple equation derived from the quadrature principle:
SNR_norm_dB = SNR_perpixel_dB + 20 * log10 [(Mpx / 12,7)^0.5]
being Mpx the pixelcount in Mpx of the camera used.
-
- Apply gaussian blur (small radius) so that each pixel becomes represented by the average exposure in its surrounding area
So you are applying a low pass filtering even before SNR calculation. To me that doesn't seem the right thing to do as it will disturb the image signal and noise relationship in the actual data.
- Match those values obtained for each pixel against the SNR response, providing a continuous map of SNR of the RAW file
Which SNR respone? I don't understand.
SNR correction because of resizing is also easy, just needs to use a corrected SNR response. For example to normalise SNR to be comparable to Canon 5D's 12Mpx, I applied this simple equation derived from the quadrature principle:
SNR_norm_dB = SNR_perpixel_dB + 20 * log10 [(Mpx / 12,7)^0.5]
being Mpx the pixelcount in Mpx of the camera used.
I don't think that is technically the correct formula to be used. Remember, I asked where is the "signal" in the normalization process. Can you please comment where or what is the signal component in this equation?
Sincerely,
Joofa
-
Yes, you are correct, and that is a way. And, I agree with that. However, as I said I don't know how many people with appreciate 10, 20, 30, ..., SNRs figures attached to a single image.
As far as photography is concerned, it is probably of no practical interest. In scientific or medical imaging, it is very important. When you image multiple point sources, such as stars, they will all have different SNR. The SNR gives you a good idea of the precision of your measure (not necessarily of its accuracy). There's a treshold under which your data can't reasonably be significant. In other fields, such as medical imaging, you'll want to know how little you can spray the patient with potentially damaging x-ray photons and still get a significant image of the intended target.
But again, the core issue is simple. There is an optimal way to expose a sensor: maximize the signal without clipping it (and ideally keep it in the sensor's linear zone). The two problems ETTR addresses are 1) some cameras have sensors whose dynamic range is too small to cover a scene as we'd like to see it. 2) those cameras typically implement ETTL - they expose to the left of the ideal exposure to protect most users in most practical scenarios. ETTR simply slides the dynamic range window where it should be. With a camera with a wide enough dynamic range window, you don't care. Whether it increases the number of levels, whether we perceive those levels, whether Michael's explanation was 100% accurate, etc... is essentially irrelevant because there is only one exposure that is ideal for the sensor, the one that puts the maximum of photons in the best zone.
-
- Apply gaussian blur (small radius) so that each pixel becomes represented by the average exposure in its surrounding area
Why is it that spatial averaging is needed again?
If you want to define a spatial neighborhood based on brightness similarity (and spatial proximity), you might want to do something like bilateral filtering.
-h
-
So you are applying a low pass filtering even before SNR calculation. To me that doesn't seem the right thing to do as it will disturb the image signal and noise relationship in the actual data.
Which SNR respone? I don't understand.
The SNR is not obtained from the image, but from the SNR response (i.e. the SNR curves that provide the SNR in that sensor for each exposure value).
The only reason for the blurring is to obtain the average exposure in the pixel's surrounding area. A single pixel's SNR cannot be measured unless you have a before/after noise addition pair of images, and even in that case is irrelevant since the observer doesn't perceive individual pixel's SNR, but local SNR. The local averaging (blurring) provides this average perceived exposure that we just need to translate to the SNR curve of the sensor to obtain the average local SNR.
I don't think that is technically the correct formula to be used. Remember, I asked where is the "signal" in the normalization process. Can you please comment where or what is the signal component in this equation?
That formula is the extrapolation of the 4 pixels binning formula, very well explained in Emil's article.
Joining 4 pixels in one improves SNR by 2, the scaling factor:
(4/1)^0,5=2
The scaling factor from a X Mpx camera to a 12,7Mpx camera is:
(X/12,7)^0,5
The formula I showed is just this scaling in dB.
Perhaps the tutorial (http://www.guillermoluijk.com/tutorial/noisedr/index.htm) about measuring the SNR curves can help to understand the process.
-
The SNR is not obtained from the image, but from the SNR response (i.e. the SNR curves that provide the SNR in that sensor for each exposure value).
Were those curves derived from real images or flat patches?
That formula is the extrapolation of the 4 pixels binning formula, very well explained in Emil's article
Joining 4 pixels in one improves SNR by 2, the scaling factor:
(4/1)^0,5=2
The scaling factor from a X Mpx camera to a 12,7Mpx camera is:
(X/12,7)^0,5
The formula I showed is just this scaling in dB.
Technically, Emil is incorrect here: This formula will work (1) only for a flat patch, and not for a real image, and (2) for downsampling by binning, which is a poor method. How will the formula change if I use Lanczos that is what many (most?) people will prefer for downsampling as opposed to binning?
Sincerely,
Joofa
-
Were those curves derived from real images or flat patches?
From both, real images of flat patches as explained in the tutorial.
Technically, Emil is incorrect here: This formula will work (1) only for a flat patch, and not for a real image, and (2) for downsampling by binning, which is a poor method. How will the formula change if I use Lanczos that is what many (most?) people will prefer for downsampling as opposed to binning?
No idea how the formula will change for other type of rescaling, I just chose this criteria because the only intention was to compare sensors. With other type of rescaling the change in SNR will be different of course, but for typical interpolation methods I'm pretty sure this simple model works well.
Regards
-
The only reason for the blurring is to obtain the average exposure in the pixel's surrounding area. A single pixel's SNR cannot be measured unless you have a before/after noise addition pair of images, and even in that case is irrelevant since the observer doesn't perceive individual pixel's SNR, but local SNR. The local averaging (blurring) provides this average perceived exposure that we just need to translate to the SNR curve of the sensor to obtain the average local SNR.
So there are two reasons:
1. To avoid noise affecting the (effectively) estimation of signal in isolation
2. To better emulate the human perception
-
Guillermo, I must say that I am very pleased at your measured, sensible, and polite responses. Thanks for all the discussion.
From both, real images of flat patches.
You mean an image acquired of a flat patch?
No idea how the formula will change for other type of rescaling, I just chose this criteria because the only intention was to compare sensors. With other type of rescaling the change in SNR will be different of course, but for typical interpolation methods I'm pretty sure this model works well.
Okay that is fine, as long as you have an appreciation of the formula changing due to a change in the resampling method.
Sincerely,
Joofa
-
So there are two reasons:
1. To avoid noise affecting the (effectively) estimation of signal in isolation
2. To better emulate the human perception
Correct!.
whether Michael's explanation was 100% accurate, etc... is essentially irrelevant because there is only one exposure that is ideal for the sensor, the one that puts the maximum of photons in the best zone.
I dont' fully agree. Reading Michael's article, one might think there are two practical advantages derived from ETTR:
1. Less visible noise
2. More levels
So even if noise is not a problem in your application at a conventional exposure, one could insist in ETTR thinking he will get some other advantage. But this is not true since those extra levels don't mean any practical benefit, so the conclusion is that if noise is not an issue in your application, insisting in ETTR is a complete waste of time.
-
You mean an image acquired of a flat patch?
Yes. A patch only means obtaining a pair (exposure, SNR). Several patches allow to plot the SNR curve.
(http://www.guillermoluijk.com/tutorial/noisedr/rawnalyze.gif)
Signal in the patch is the average value on it and hence the patch exposure. Noise in the patch is the patch's stdDev. Having S and N, we have the desired (S, SNR).
The blurring of real images was just to obtain the S value (local average exposure). The SNR value is then obtained from the curve for the X-axis value indicated by the previous S.
-
Yes. A patch only means obtaining a pair (exposure, SNR). Several patches allow to plot the SNR curve.
Are we back to square one? A real image (of your cat, dog, cat, people, cat, landscape, cat, ..., cat) is not always easy to segment into patches as you have shown in the image in your example above. Do you agree with this statement?
Sincerely,
Joofa
-
Correct!.
At very large signal levels, the SNR should be large. Then noise should not affect the estimation of signal (and thereby SNR). The likelihood that a "code 20" signal is changed into a code 16211 by noise should hopefully be very, very low.
One problem with including perception: if we (primarily) sense the world limited to e.g. 1 MP, then any camera/raw converter that does crude NR by downsampling would appear to have better SNR. Is this sensible?
So even if noise is not a problem in your application at a conventional exposure, one could insist in ETTR thinking he will get some other advantage. But this is not true: if noise is not an issue in your application, insisting in ETTR is a complete waste of time.
Even if noise is not considered a problem at the time of capture, I think that some would prefer to capture it as accurately as possible (who knows what post-processing will be applied many years into the future).
Is this based on self-dithering properties of noise, or some other post of yours that I did not comprehend?
-h
-
Are we back to square one? A real image (of your cat, dog, cat, people, cat, landscape, cat, ..., cat) is not always easy to segment into patches as you have shown in the image in your example above. Do you agree with this statement?
Of course, but we don't need to segment it in patches, just would calculate each pixel's surrounding exposure (that's the blur filter).
Another option is to segment the image in wide EV zones, and consider a SNR for each whole EV area. But again we need some kind of averaging; otherwise noise and textures could fool us:
(http://www.guillermoluijk.com/tutorial/histogrammar/zonasluma.gif)
This image was noiseless, since it was produced with a blending of several exposures, but still textures in the ground make it difficult to clearly differentiate the EV borders.
-
Are we back to square one? A real image (of your cat, dog, cat, people, cat, landscape, cat, ..., cat) is not always easy to segment into patches as you have shown in the image in your example above. Do you agree with this statement?
Joofa,
Your past few posts have been argumentative and unhelpful. One can not give a SNR for an entire image. Each area will have a different SNR according to the luminance of the area being studied. The highlights will have a high SNR and the shadows a low SNR. Look at the SNR plot by DXO for the D7000 for base ISO. If the image is fully exposed to the right, the highlights will have a SNR of about 45 dB. The deep shadows will have a lower SNR.
Regards,
Bill
-
Even if noise is not considered a problem at the time of capture, I think that some would prefer to capture it as accurately as possible (who knows what post-processing will be applied many years into the future).
But you must think ETTR is not for free. You need longer exposure times (that may cause a not so sharp image), or wider apertures (losing DOF) or higher ISO (clipping highlights information). In general, once noise requirements are met, I prefer to use the fastest shutter I can afford to optimise sharpness.
-
Joofa,
Your past few posts have been argumentative and unhelpful.
Sorry Bill if you thought so.
One can not give a SNR for an entire image. Each area will have a different SNR according to the luminance of the area being studied. The highlights will have a high SNR and the shadows a low SNR. Look at the SNR plot by DXO for the D7000 for base ISO. If the image is fully exposed to the right, the highlights will have a SNR of about 45 dB. The deep shadows will have a lower SNR.
I think I already answered this here:
http://www.luminous-landscape.com/forum/index.php?topic=56947.msg464138#msg464138
Sincerely,
Joofa
-
But you must think ETTR is not for free. You need longer exposure times (that may cause a not so sharp image), or wider apertures (losing DOF) or higher ISO (clipping highlights information). In general, once noise requirements are met, I prefer to use the fastest shutter I can afford to optimise sharpness.
So ETTR it is a sort-of single-sided requirement: if circumstances allow, increase exposure until nearly clipping. If not, do whatever compromise between motion-blur, DOF and noise that satisfy your artistic requirements.
There are cases where the problem is small aperture due to bright sunlight and unwanted diffraction, perhaps using flash for evening out the light (and limiting the minimum exposure time), and the lack of ND filtering or lower sensitivity ISO modes. In that case, knowing exactly how "hot" you can expose the sensor using ETTR-inspired techniques and knowing that it will only gain the SNR seems like a good thing to know.
-h
-
So even if noise is not a problem in your application at a conventional exposure, one could insist in ETTR thinking he will get some other advantage. But this is not true since those extra levels don't mean any practical benefit, so the conclusion is that if noise is not an issue in your application, insisting in ETTR is a complete waste of time.
On the whole I agree - but saying "if noise is not an issue" essentially means that the DR of the scene as you want to capture it fits nicely in the DR window of your camera. And if it fits, but at the bottom, fitting it at the top will in most cases give a bit more latitude for post processing. Also agree on the cost of ETTR, although I found out with practice that using ETTR-Light (overexposing slightly and systematically) with my Canon cameras is 99.9% of the time effective. It reverses the effect of the defensive defaults chosen by the Canon engineers.
-
Another option is to segment the image in wide EV zones, and consider a SNR for each whole EV area. But again we need some kind of averaging; otherwise noise and textures could fool us:
That is one way to proceed and fine with me. A point of the whole exercise was how to figure out that "some kind of averaging", and how to separate noise from texture, as you have also identified. That is not always an easy task to do.
Sincerely,
Joofa
-
Joofa,
Your past few posts have been argumentative and unhelpful.
I wouldn't say so. He raises some valid points. While Emil's work is extremely impressive and beyond reproach, some of its results, especially in the normalization of SNR vs Resolution area, should not be extended to extremes by means of black boxes (resize operations in software for example, which are not equivalent to hardware binning). But it is again a long topic of little practical interest for photography.
-
But you must think ETTR is not for free. You need longer exposure times (that may cause a not so sharp image), or wider apertures (losing DOF) or higher ISO (clipping highlights information). In general, once noise requirements are met, I prefer to use the fastest shutter I can afford to optimise sharpness.
Just to be clear, if one is already shooting on a tripod (useful to get maximum image sharpness) the penalty for slowing down the shutter speed isn't great. Still a factor for subject motion blur, yes...but not for camera shake. If you are shooting hand held, yes, the focal length and shutter speed will be a factor that must be weighed and decided about...it's far easier to use noise reduction to mitigate shot noise than try to use a sharpening effect to correct for camera shake.
There are a lot of factors that must be evaluated when making a capture...but the science still shows that more photons is always better when it comes to noise. When to use ETTR and when it'll help the image quality is for the photographer to decide, scene by scene and image by image. Above all else, take the shot...don't let all the various factors that must be considered keep you from pressing the shutter. Take the shot...
-
There are a lot of factors that must be evaluated when making a capture...but the science still shows that more photons is always better when it comes to noise. When to use ETTR and when it'll help the image quality is for the photographer to decide, scene by scene and image by image. Above all else, take the shot...don't let all the various factors that must be considered keep you from pressing the shutter. Take the shot...
Right on! TTFP....Take The F***ing Picture!
-
I wouldn't say so. He raises some valid points. While Emil's work is extremely impressive and beyond reproach, some of its results, especially in the normalization of SNR vs Resolution area, should not be extended to extremes by means of black boxes (resize operations in software for example, which are not equivalent to hardware binning). But it is again a long topic of little practical interest for photography.
As Emil correctly points out, software averaging of 4 pixels gives twice the SNR of using only one pixel, since noise adds in quadrature. Hardware binning is frequently used in scientific applications with monochrome sensors, but is much more difficult to do with Bayer array sensors. The only color cameras for general photographic use AFAIK that use hardware binning are the Phase One Sensor Plus series. Hardware binning of 4 pixels gives 4 times the SNR, since only one read noise is involved, whereas 4 read noises occur with software pixel averaging. See the explanation by an engineer on the Phase One site (http://www.phaseone.com/en/Digital-Backs/IQ180/IQ180-Tutorials.aspx) (click on the IQ Sensor+ tab).
Noise averaging is of great importance in photography. If one holds sensor size constant and doubles the pixel count, the pixels are smaller and have a lower SNR per pixel, but the overall SNR will not be affected that much. When comparing cameras with different pixel counts for a given print size, some type of normalization must be performed as in the DXO screen and print data.
Regards,
Bill
-
As Emil correctly points out, software averaging of 4 pixels gives twice the SNR of using only one pixel, since noise adds in quadrature.
Unfortunately, this statement is technically incorrect in general in the context of images, as pointed out several times.What you said will only work if the signal is constant (flat patch). Think about it like this: even if we don't worry about how noise is being added, when you add pixels you are changing (usually blurring) the pixels also, so the signal has changed. Recall, SNR is signal to noise ratio, so SNR changes differently than "4 pixels gives twice the SNR of using only one pixel". I would invite Emil to do calculations himself to verify this fact rather than theoretical arguments.
Sincerely,
Joofa
-
Unfortunately, this statement is technically incorrect in general in the context of images, as pointed out several times.What you said will only work if the signal is constant (flat patch). Think about it like this: even if we don't worry about how noise is being added, when you add pixels you are changing (usually blurring) the pixels also, so the signal has changed. Recall, SNR is signal to noise ratio, so SNR changes differently than "4 pixels gives twice the SNR of using only one pixel". I would invite Emil to do calculations himself to verify this fact rather than theoretical arguments.
I will await Emil's response, since he is more technically adept than myself. In the meantime, what are your calculations?
Regards,
Bill
-
Unfortunately, this statement is technically incorrect in general in the context of images, as pointed out several times.What you said will only work if the signal is constant (flat patch). Think about it like this: even if we don't worry about how noise is being added, when you add pixels you are changing (usually blurring) the pixels also, so the signal has changed. Recall, SNR is signal to noise ratio, so SNR changes differently than "4 pixels gives twice the SNR of using only one pixel". I would invite Emil to do calculations himself to verify this fact rather than theoretical arguments.
While awaiting Emil's analysis (presuming he takes the trouble to reply to your post), the DXO normalization procedure (http://www.dxomark.com/index.php/Publications/DxOMark-Insights/Detailed-computation-of-DxOMark-Sensor-normalization) is of interest. The normalized SNR equation is:
SNRnorm = SNR + 20 * log10 (sqrt[N/N0]),
where N0 is the original number of pixels, N is the number of pixels for the sensor with the higher pixel count, and SNR is the original SNR.
If we average 4 pixels into one, the formula shows that the SNR increases by 6.02 dB or 1 stop, in agreement with Emil's figure. I think DXO is using flat patches to derive their figures. As you suggest, in real world use with demosaiced images, the SNR may be somewhat less than the theoretical value. The DXO engineer (http://www.phaseone.com/en/Digital-Backs/IQ180/IQ180-Tutorials.aspx) also states that 4:1 binning outside the sensor hardware doubles the SNR whereas hardware binning quadruples the SNR. What are your figures?
Regards,
Bill
-
the DXO normalization procedure (http://www.dxomark.com/index.php/Publications/DxOMark-Insights/Detailed-computation-of-DxOMark-Sensor-normalization) is of interest. The normalized SNR equation is:
SNRnorm = SNR + 20 * log10 (sqrt[N/N0]),
where N0 is the original number of pixels, N is the number of pixels for the sensor with the higher pixel count, and SNR is the original SNR.
If we average 4 pixels into one, the formula shows that the SNR increases by 6.02 dB or 1 stop, in agreement with Emil's figure. I think DXO is using flat patches to derive their figures. As you suggest, in real world use with demosaiced images, the SNR may be somewhat less than the theoretical value.
I think the 'patches discusion' is going farer than it really deserves. Averaging 4 pixels of the same value + their individually added noise improves SNR by 2, I think we all agree here. If the signal is different on each pixel, the noise will also be different, so SNR will be different on each pixel. If each source pixel has its own SNR, it's nonsense to look for 'SNR improves by a factor of X', since source SNR is not unique.
But once defined, the 'patch model' can be extended to any real world case with the purpose of SNR normalisation. It will always be an approximation, but a valid approximation to make noise performance from different sensors comparable. It is not intended to quantify SNR on any real picture of our cat.
That DxOMark's formula was already posted in the thread as derived from quadrature noise addition:
SNR_norm_dB = SNR_perpixel_dB + 20 * log10 [(Mpx / 12,7)^0.5]
Regards
-
While awaiting Emil's analysis (presuming he takes the trouble to reply to your post), the DXO normalization procedure (http://www.dxomark.com/index.php/Publications/DxOMark-Insights/Detailed-computation-of-DxOMark-Sensor-normalization) is of interest. The normalized SNR equation is:
SNRnorm = SNR + 20 * log10 (sqrt[N/N0]),
where N0 is the original number of pixels, N is the number of pixels for the sensor with the higher pixel count, and SNR is the original SNR.
If we average 4 pixels into one, the formula shows that the SNR increases by 6.02 dB or 1 stop, in agreement with Emil's figure. I think DXO is using flat patches to derive their figures. As you suggest, in real world use with demosaiced images, the SNR may be somewhat less than the theoretical value. The DXO engineer also states that 4:1 binning will give 1 one stop improvement in SNR for the DXO engineer (http://www.phaseone.com/en/Digital-Backs/IQ180/IQ180-Tutorials.aspx) also states that binning outside the sensor doubles the SNR whereas hardware binning quadruples the SNR. What are your figures?
Hi Bill,
joofa's numbers are the same, he is just acting a bit anal about the fact that most real life images are dominated by non-uniform (image detail) patches, so it's hard to calculate average noise (the amount of noise is different for each pixel where the signal is different as well).
Even with uniform areas, a non-binning type of averaging further complicates the noise averaging statistics, because the downsampling filters are not uniform over their support width (e.g. Lanczos window). He conveniently does not mention the possibility (although not practical with subject motion) to average mutiple exposures of the same scene, as is e.g. done routinely by astrophotographers. That way each pixel gets averaged as expected and noise is reduced in quadrature, regardless of its ajacent pixel's signal level.
Cheers,
Bart
-
The point is that noise acts to dither the levels. Suppose the true signal is some value X between zero and one, over a patch of the image (we are going to ignore natural scene variation for the purpose of answering your question). Suppose the noise is of strength N, for example let N be one level. The noise adds a random number roughly between -N and N to X, so that the pixel wants to record some number between X-N and X+N. Of course, the resulting signal plus noise is digitized so the output is either 0 or 1; if the noise is random (uncorrelated from pixel to pixel), the value of X is reflected in the percentage of 1's vs 0's in the patch -- a fraction X of the pixels will be 1 and the rest 0. If we average the levels over a large enough patch, we recover the original signal, even though each individual pixel only recorded 0 or 1.
This is the basic idea that allows one to trade resolution for noise -- and why downsampled images look less noisy. Note also that while the average is more finely graded than steps of one, that doesn't mean we buy anything by making individual pixels record values more finely spaced than the level of noise, because the individual values are jumping around randomly by an amount between -N and N.
Thanks for attempting the explanation, Emil, even though I have to admit it is not all totally clear.
Nevertheless, I can see with my own eyes that noisy images after downsampling exhibit less noise. A similar effect can be achieved simply by viewing the image from a greater distance.
What still puzzles me is that a 14 bit A/D converter in a camera with a good dynamice range, such as the D7000, will not provide any IQ advantages in the deepest shadows, compared with 12 bit.
However, I can understand this might be the case with a Canon camera which doesn't go beyond 12 stops of DR.
I've gone back to the RAW files and created crops of the 12th, 13th and 14th stops, which show a clear improvement in IQ in the 12th stop.
Now, you seem to be claiming that such improvement in that 12th stop is entirely due to the greater quantity of light resulting in an improved SNR (1/500th sec as opposed to 1/2000th for the 14th stop).
Of course, I am able to appreciate that 4x the number of photons will improve the SNR and the IQ as a result.
I'm just a bit skeptical that the image quality in that 12th stop, as shown in my attached image, could be achieved with just the one level per channel of a 12 bit converter, as opposed to the 4 levels per channel of the 14 bit converter.
Normally, I would just repeat the test in 12 bit and 14 bit modes. If I see a difference at the same very low exposures, then there's a difference, depite any theory from an eminent Physicist.
Unfortunately, my D7000 is in for repair due to its inability to autofocus in cold weather. It looks as though I won't get it back for a while because it is now Spring in Australia, and the Nikon repair agent is unable to duplicate the problem in the artificially cold environments he has created.
Tell me now whether I can expect to see the same image quality in that 12th stop with the D7000 set to 12 bit mode, and when I get the camera back, or a new replacement, I'll carry out the tests and tell you whether you are right or wrong. :)
-
I think the 'patches discusion' is going farer than it really deserves. Averaging 4 pixels of the same value + their individually added noise improves SNR by 2, I think we all agree here. If the signal is different on each pixel, the noise will also be different, so SNR will be different on each pixel. If each source pixel has its own SNR, it's nonsense to look for 'SNR improves by a factor of X', since source SNR is not unique.
Quite true. Also, we should remember that the SNR for each color of the Bayer array will be different. Figures are usually given for the green channels, since they should be near saturation with proper exposure. However, the red and blue channels will be below saturation according to the white balance multipliers. Nonetheless, their noise will also be reduced by a like factor when adding in quadrature.
Regards,
Bill
-
What are your figures?
http://forums.dpreview.com/forums/read.asp?forum=1018&message=32867870
And, remember, that is just one way of doing it. In fact, I have changed my own approach somewhat from the above links that is years older.
joofa's numbers are the same, he is just acting a bit anal about the fact that most real life images are dominated by non-uniform (image detail) patches, so it's hard to calculate average noise (the amount of noise is different for each pixel where the signal is different as well).
I don't know how can you claim the numbers are the same, without even knowing what I did. Kindly see the link posted above.
Even with uniform areas, a non-binning type of averaging further complicates the noise averaging statistics, because the downsampling filters are not uniform over their support width (e.g. Lanczos window). He conveniently does not mention the possibility (although not practical with subject motion) to average mutiple exposures of the same scene, as is e.g. done routinely by astrophotographers. That way each pixel gets averaged as expected and noise is reduced in quadrature, regardless of its ajacent pixel's signal level.
I don't think you realize that there is more to it than quadrature. Noise still adds in quadrature, but that is not the end of story. Again see the link above.
Sincerely,
Joofa
-
The DXO engineer (http://www.phaseone.com/en/Digital-Backs/IQ180/IQ180-Tutorials.aspx) also states that 4:1 binning outside the sensor hardware doubles the SNR whereas hardware binning quadruples the SNR. What are your figures?
Regards,
Bill
But using digital processing, the choice of averaging kernel is without limits, including negative coefficients, space-variant, signal-dependent (non-linear) manually guided processing compared to the simple box-car filter that binning even non-bayer sensels. This might (or might not) be enough to offset the initial advantage of sensor binning.
-h
-
I think the 'patches discusion' is going farer than it really deserves.
What is the reason that patches are being discussed? Just curiosity, or is there some real-world photography application being limited by this approach?
-h
-
http://forums.dpreview.com/forums/read.asp?forum=1018&message=32867870
And, remember, that is just one way of doing it. In fact, I have changed my own approach somewhat from the above links that is years older.
I don't know how can you claim the numbers are the same, without even knowing what I did.
And therein lies the problem with you. You are not very clear in what you base your assumptions on (so one can only assume common statistical principles which apply to us all), yet you hint at outcomes that deviate from common experience. You then don't explain but refer to another website where you also cause confusion by not explaining that you use a model that assumes correlation between sensels. Why? Nobody but you knows, until the info is tortured out of you. Why? And then (above) you add that the models you use in the discussion on that website are not the models you now use. Well thanks for creating yet more confusion. Was the earlier model not correct? What are you using now? Why are you using a correlation model in the first place, any literature/references/experiments you want to share that other can independently confirm?
No wonder Emil has given up on 'discussing' with you.
Noise still adds in quadrature, but that is not the end of story.
More riddles. I'm not even going to ask...
Sincere cheers,
Bart
-
I will await Emil's response, since he is more technically adept than myself. In the meantime, what are your calculations?
Emil is not going to follow joofa's red herrings. This was a thread about ETTR; I am happy to discuss that topic, though I suspect it has been covered amply already. If the subject is to be changed, how about mass in general relativity? Did you know that there is no precise definition of local mass in Einstein's theory? Somehow, it doesn't keep my bathroom scale from working...
-
any literature/references/experiments you want to share that other can independently confirm?
At DPReview I have gone over this topic several times with details of what I'm doing. You are welcome to browse my messages there. But in summary, I just used elementary stochastic processes as applied in electrical engineering.
No wonder Emil has given up on 'discussing' with you.
I think signal processing is not Emil's forte ;D.
Sincerely,
Joofa
-
At DPReview I have gone over this topic several times with details of what I'm doing. You are welcome to browse my messages there. But in summary, I just used elementary stochastic processes as applied in electrical engineering.
I think signal processing is not Emil's forte ;D.
I can see why you are not pleased with Emil. In the thread on DPReview which you referenced, he was apparently aware of your thesis but chose to ignore it. :)
Regards,
Bill
-
I can see why you are not pleased with Emil. In the thread on DPReview which you referenced, he was apparently aware of your thesis but chose to ignore it. :)
Bill, I have a great deal of respect for you. Lets not make this personal. It's Emil's prerogative to ignore it if he wants to and that is fine with me. I stand by my assertions and anybody who doubts them can work out the numbers themselves.
Sincerely,
Joofa
-
Before anyone gets further bent out of shape it appears Joofa is simply saying in a practical sense that the basic 6.02 dB figure for resizing by a factor of two is complicated by two issues.
First, it is rare for anyone to use a rectangular kernel when resizing and practical kernels will not necessarily produce the same change in SNR. There are a variety of reasons for this, but one would be the differing spatial frequency distribution of the scene detail and noise combined with a kernel with a non-uniform frequency response.
Second, since the noise distribution is amplitude dependent in a real scene the statistics will be more complicated than a simple model based on uniform patches. This will also cause a deviation from the 6.02 dB figure.
Of what practical use this point is to the ETTR discussion in this thread I'm not sure. It seems like a thread hijack of thread hijack. Regardless, he's not off base in stating the simple 6.02 dB model used by DxO is in fact a simplification complicated by real world images and real world resizing. No need to jump down his throat about it.
If you follow Joofa's posts here and elsewhere you'll know he likes to ask questions instead of answer them and always leaves dangling caveats to everything he says. If you spend any time in academia you will encounter personalities who seem to think obfuscation makes them appear erudite. Joofa is one of those souls who hasn't figured out clear communication is a better standard of competence. You'll need to cut him some slack if you want to figure out what he's trying to say, and usually you'll find he isn't too far off base. Whether it is worth parsing his rhetorical methods is for the reader to decide.
Ken
-
BTW, I've just finished a nice book that should be mandatory reading for anyone taking part in longish threads on the Internet ;)
http://www.amazon.com/Being-Wrong-Adventures-Margin-Error/dp/0061176052/
-
BTW, I've just finished a nice book that should be mandatory reading for anyone taking part in longish threads on the Internet ;)
http://www.amazon.com/Being-Wrong-Adventures-Margin-Error/dp/0061176052/
But this ONE (http://www.amazon.com/Extraordinary-Popular-Delusions-Madness-Crowds/dp/1463740514/ref=sr_1_1?s=books&ie=UTF8&qid=1314832985&sr=1-1) is far better and more in keeping with what is going on!
-
EPDM ? i thought that was a kind of synthetic rubber ;D
Frank
-
In case anyone is still interested in the differences between 12 bit and 14 bit A/D conversions, I've done a few tests with my D7000 of the same target shown before.
I'm surprised that there is not a greater difference in the 12th stop and 11th stop. The differences are subtle. The 12 bit shots exhibit slightly greater grain and a degree of clipping of blacks that no adjustment in ACR can completely correct.
However, in the 13th and 14th stops of DR the differences smack one in the face.
One could argue that such differences are of no consequence because they are apparent only in the deepest shadows which one might want to deliberately clip. However, it seems to me if one is underexposing say 5 stops at ISO 100 as an alternative to using ISO 3200, one wants to be able to retrieve as much detail as possible from the shadows which will appear totally black before exposure compensation.
I'm going to try posting a couple of comparisons of the 13th and 14th stops of DR, but the jpeg engine seems to have difficulty compressing noise and the file sizes are quite large. So, if you don't see the images, don't be surprised. I may have to post them one at a time.
-
In case anyone is still interested in the differences between 12 bit and 14 bit A/D conversions, I've done a few tests with my D7000 of the same target shown before.
I'm surprised that there is not a greater difference in the 12th stop and 11th stop. The differences are subtle. The 12 bit shots exhibit slightly greater grain and a degree of clipping of blacks that no adjustment in ACR can completely correct.
However, in the 13th and 14th stops of DR the differences smack one in the face.
Since a 12 bit linear file can only encode 12 stops of DR, it is confusing to try to extract 14 bits of DR from a 12 bit file.
Regards.
Bill
-
Since a 12 bit linear file can only encode 12 stops of DR, it is confusing to try to extract 14 bits of DR from a 12 bit file.
Regards.
Bill
Indeed it is. But Emil's point, as I understood it, was that noise would prevent any advantage of an increase in the number of levels beyond 12 bit, and that 14 bits served no purpose other than a possible slight increase in the accuracy of the conversion.
Perhaps it would be clearer to state, if the DR of the camera is no greater than 12 stops, then more levels that 12 bits afford, is of little use.
However if the camera has a DR of more than 12 stops, then more than 12 bits is of use.
If one wishes to engage in pixel peeping, 14 bits does provided a pixel-peeping IQ advantage in the 11th and 12th stops, with the D7k. However, I doubt that would be the case with lesser cameras, DR-wise.
-
Indeed it is. But Emil's point, as I understood it, was that noise would prevent any advantage of an increase in the number of levels beyond 12 bit, and that 14 bits served no purpose other than a possible slight increase in the accuracy of the conversion.
The D7000 is such a low noise camera that less than 14 bits are not sufficient to properly encode the deep shadows information at base ISO. But this is compatible with the fact that as long as noise dithers posterization, having more levels (i.e. having more bits) is useless.
I developed a RAW file from the Pentax K5 (same sensor as D7000), emulating a 12-bit RAW file (by decimating the RAW data from 14-bit to 12-bit prior to RAW development), and posterization began to show up:
14-bit: noise dithers posterization
(http://www.guillermoluijk.com/article/rawbits/k5_1.jpg)
12-bit: posterization becomes visible
(http://www.guillermoluijk.com/article/rawbits/k5_2.jpg)
However, doing the same test on Canon 40D's RAW files, there was no loss of IQ using 12-bit. That camera is noisy enough to make 14 bits unnecesary.
Regards
-
Indeed it is. But Emil's point, as I understood it, was that noise would prevent any advantage of an increase in the number of levels beyond 12 bit, and that 14 bits served no purpose other than a possible slight increase in the accuracy of the conversion.
Perhaps it would be clearer to state, if the DR of the camera is no greater than 12 stops, then more levels that 12 bits afford, is of little use.
However if the camera has a DR of more than 12 stops, then more than 12 bits is of use.
If one wishes to engage in pixel peeping, 14 bits does provided a pixel-peeping IQ advantage in the 11th and 12th stops, with the D7k. However, I doubt that would be the case with lesser cameras, DR-wise.
I agree. I don't know if Emil has studied the D7000 or the D3x. In his treatise (http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/noise-p3.html), he does state that the D3 comes close to warranting a 13th bit, so I would think by extrapolation that the D7000 could make use of more than 12 bits.
Regards,
Bill
-
Indeed it is. But Emil's point, as I understood it, was that noise would prevent any advantage of an increase in the number of levels beyond 12 bit, and that 14 bits served no purpose other than a possible slight increase in the accuracy of the conversion.
Perhaps it would be clearer to state, if the DR of the camera is no greater than 12 stops, then more levels that 12 bits afford, is of little use.
However if the camera has a DR of more than 12 stops, then more than 12 bits is of use.
If one wishes to engage in pixel peeping, 14 bits does provided a pixel-peeping IQ advantage in the 11th and 12th stops, with the D7k. However, I doubt that would be the case with lesser cameras, DR-wise.
Noise will effectively dither, when the noise is sufficiently large relative to the quantization step. If you expose to mostly only populate the 13th and 14th bit of data, and then truncate as you have to 12 bits, then of course it's going to look like garbage.
-
I developed a RAW file from the Pentax K5 (same sensor as D7000), emulating a 12-bit RAW file (by decimating the RAW data from 14-bit to 12-bit prior to RAW development), and posterization began to show up:
[...]
However, doing the same test on Canon 40D's RAW files, there was no loss of IQ using 12-bit. That camera is noisy enough to make 14 bits unnecesary.
Hi Guillermo,
When you say unnecessary, I wonder if that is true under all circumstances. Maybe 'hardly necessary' is closer to the truth? Have you for instance tried averaging, say 4 or 16, different exposures (as is routinely done in astophotography) and then analyse the resulting noise level? If there is truely no benefit for, or information in, the 13th and 14th bit then you would be correct and both would have identical noise levels and posterization ...
Cheers,
Bart
-
However, doing the same test on Canon 40D's RAW files, there was no loss of IQ using 12-bit. That camera is noisy enough to make 14 bits unnecesary.
I suspect that would be true. I tried similar tests with my D700 whilst my D7k was in for repair, and was very surprised to find virtually no image quality difference in the 11th stop, between 12 bit and 14 bit, except for a slight additional graininess in the 12 bit image, almost like the difference between an ISO 200 film and an ISO 400 film, but not as great.
I admit I was expecting to see a greater difference.
However, the D7k is a different kettle of fish. Not only is that slightly greater graininess apparent in the 12 bit shots, in the 11th and 12th stops also, but there seems to be an unavoidable clipping of the blacks in 12 bit mode. Such clipping is not apparent in the 14 bit shots with the D7k.
It goes without saying that I've applied no sharpening nor noise reduction in the following ACR conversions. Tone curve is linear, blacks zero and contrast either 0 or -50, depending on the image.
-
Noise will effectively dither, when the noise is sufficiently large relative to the quantization step. If you expose to mostly only populate the 13th and 14th bit of data, and then truncate as you have to 12 bits, then of course it's going to look like garbage.
Emil,
I admit those images I've presented, that are extremely underexposed, look like garbage.
The issue for me is, do the 12 bit images look even more like garbage than the 14 bit images?
In a real scene with a correct exposure, only the deepest shadows would look like a crop of my underexposed shots, after selecting and raising such shadows in the real scene, in order to see the detail.
I believe my methodology is sound. However, I agree with you that ETTR is mainly about SNR and little about the number of levels, considering that most cameras probably provide an excess of levels in relation to the noise floor.
-
Back in the 1960´s, when I was a wee lad, audiophiles were the serious nerds, competing to build a better sound system. The really serious ones did not bother to listen to their systems. Instead they connected an oscilloscope to the output of their latest amplifier, and sat entranced by the pureness of the waveforms displayed on their little green screens.
This discussion seems dangerously close to veering in that direction. Or as Yogi Berra put it: "It´s déjá vu all over again!"
-
Back in the 1960´s, when I was a wee lad, audiophiles were the serious nerds, competing to build a better sound system. The really serious ones did not bother to listen to their systems. Instead they connected an oscilloscope to the output of their latest amplifier, and sat entranced by the pureness of the waveforms displayed on their little green screens.
This discussion seems dangerously close to veering in that direction. Or as Yogi Berra put it: "It´s déjá vu all over again!"
Ooh, great idea! I'll have to get an oscilloscope right away... ;D
-
Back in the 1960´s, when I was a wee lad, audiophiles were the serious nerds, competing to build a better sound system. The really serious ones did not bother to listen to their systems. Instead they connected an oscilloscope to the output of their latest amplifier, and sat entranced by the pureness of the waveforms displayed on their little green screens.
This discussion seems dangerously close to veering in that direction. Or as Yogi Berra put it: "It´s déjá vu all over again!"
I think I still have an ancient oscilloscope in my basement somewhere. If any of you geeks want to come and get it, you can have it. Of course you might have to hunt up some new tubes (yes, tubes) for it.
-
Back in the 1960´s, when I was a wee lad, audiophiles were the serious nerds, competing to build a better sound system. The really serious ones did not bother to listen to their systems. Instead they connected an oscilloscope to the output of their latest amplifier, and sat entranced by the pureness of the waveforms displayed on their little green screens.
This discussion seems dangerously close to veering in that direction. Or as Yogi Berra put it: "It´s déjá vu all over again!"
I am surprised that the.....you don't have to read it .....reply hasn't surfaced. At the end of the day we don't but we do read it to see if something practical has been posted. Alas no. :(
-
When you say unnecessary, I wonder if that is true under all circumstances. Maybe 'hardly necessary' is closer to the truth? Have you for instance tried averaging, say 4 or 16, different exposures (as is routinely done in astophotography) and then analyse the resulting noise level? If there is truely no benefit for, or information in, the 13th and 14th bit then you would be correct and both would have identical noise levels and posterization ...
Averaging isn't the correct operation. Accumulating is. Summing four 12-bit files gives you 14-bits of information and increases your SNR by one bit (so you could round to 13-bits without loss). Summing sixteen 12-bit files gives you 18-bits of information and increases your SNR by two bits (round to 14).
You are correct that once you accumulate or average a number of 12-bit files you may get to the point that 12-bits is no longer sufficient to keep the quantization noise sufficiently low. But this doesn't imply that 12-bits was insufficient in the individual files. If read noise was stronger than quantization noise in the individual files at 12-bits then additional bits in the individual files won't help you later on. Where you need the extra bits is after you combine the files, not before.
Ken
-
Back in the 1960´s, when I was a wee lad, audiophiles were the serious nerds, competing to build a better sound system. The really serious ones did not bother to listen to their systems. Instead they connected an oscilloscope to the output of their latest amplifier, and sat entranced by the pureness of the waveforms displayed on their little green screens.
This discussion seems dangerously close to veering in that direction. Or as Yogi Berra put it: "It´s déjá vu all over again!"
There's something wrong with this analogy. We're not using our ears to assess visual matters here. Audio is a hearing thing. Photography is a visual thing. Judging audio quality with one's eyes would be as silly as judging photographic technical quality with one's ears.
The issue of 12 bit versus 14 bit A/D conversion, in cameras, is interesting because it does seem that for many models of cameras 12 bits are sufficient, and 14 bits may be mostly an advertising gimmick.
The D7000 is probably the only camera I own where 14 bits are useful. I didn't know that until I carried out the above tests.
-
Averaging isn't the correct operation. Accumulating is.
Hi Ken,
Adding values will make the result brighter until the result is clipped by the (often) integer math word length. In the end you need to scale down (=divide by the number of images taken, AKA averaging) to get the original brightness (but with lower noise and thus more accurate). One can opt for addition, and subsequent division by less than the number of images, but that also increases the brightness of the resulting image, with potential clipping.
Summing four 12-bit files gives you 14-bits of information [...]
Unless the images were ETTR exposures with the MSB filled, or when the 12 bit images were not scaled properly on the MSB when promoting to more bits. Not scaling 12-bit input properly when converting to 14-bit or 16-bit equals a division, hence averaging takes place. It would also require an exact number of 2, 4, 8, 16 images to get the orginal brightness, which is an unnecessary restriction (15 images are almost as useful as 16, especially when the brightness doesn't drop by more than 6%).
Cheers,
Bart
-
Adding values will make the result brighter until the result is clipped by the (often) integer math word length. In the end you need to scale down (=divide by the number of images taken, AKA averaging) to get the original brightness (but with lower noise and thus more accurate). One can opt for addition, and subsequent division by less than the number of images, but that also increases the brightness of the resulting image, with potential clipping.
If you have a primitive image processing chain this is true. Black and white points are arbitrary and 64-bit integers are supported in most real processing applications (or in some 52-bits if they prefer floating point) so the reality is there is never any reason to divide or average in image processing. But yes, if you want to get back to something that will fit in an original RAW file bit depth to then pass through something like LR then you need to worry about getting back to the original integer range through division. The point still holds though, more bits in the individual capture images is irrelevant, bits in the combined output image is what matters. Forcing oneself back into the original RAW bit depth from a combination of multiple exposures can lead to quantization noise being higher than averaged read noise. This is why averaging is suboptimal compared to accumulation.
Unless the images were ETTR exposures with the MSB filled, or when the 12 bit images were not scaled properly on the MSB when promoting to more bits. Not scaling 12-bit input properly when converting to 14-bit or 16-bit equals a division, hence averaging takes place. It would also require an exact number of 2, 4, 8, 16 images to get the orginal brightness, which is an unnecessary restriction (15 images are almost as useful as 16, especially when the brightness doesn't drop by more than 6%).
Again, I was just using a simple example. The proper solution is to not round or truncate but instead just add and change your white point (and possibly black point depending on the ADC bias). In such a manner an arbitrary number of images can be handled. Again, to get back to a "standard" processing flow you'll be forced to average which is suboptimal.
Happy shooting!
Ken
EDIT: Part of that was confusing. Obviously dividing and averaging in double precision floating point would be just fine. When I said averaging was suboptimal compared to accumulation and white point adjustment I was referring to integer operations.
-
I am surprised that the.....you don't have to read it .....reply hasn't surfaced.
May be the technical crowd also realizes that the discussion on ETTR is becoming a bit futile ;D
-
Talk about obsession!!! ;D
-
Quote
EDIT: Part of that was confusing. Obviously dividing and averaging in double precision floating point would be just fine. When I said averaging was suboptimal compared to accumulation and white point adjustment I was referring to integer operations.
unquote
I think it is now obvious that each of the obsessives are now taking the mickey with each other? ;) ;D
-
Referring to the RAW files of my DR tests for the D7000, I see that the last shot in the series, beginning with a 4 secs exposure, was taken at 1/8000th exposure, which represents the 16th stop.
Surprisingly, there's still some broad detail visible through the noise. At least one can read the heading.
Hi Ray,
I recently came across this old thread and would be interested in taking a look at the NEFs (in particular DSC0037, 52, 58 and any other you may think relevant). Any chance I could trouble you to share them (dropbox, skydrive, google drive)? I have PM'd you my email address.
Thanks,
Jack
-
Hi Ray,
I recently came across this old thread and would be interested in taking a look at the NEFs (in particular DSC0037, 52, 58 and any other you may think relevant). Any chance I could trouble you to share them (dropbox, skydrive, google drive)? I have PM'd you my email address.
Thanks,
Jack
Hi Jack,
I'll have a look for them. I'm in the process of reorganizing my images so it shouldn't take too long.
Regards