Luminous Landscape Forum
Equipment & Techniques => Beginner's Questions => Topic started by: maenol on August 19, 2009, 08:15:36 am
-
Hi,
I am sorry if this is a stupid question. But being new I would like some advice. I read a lot about exposure to the right and understand that not doing so underuses the capabilities of modern sensors. However, what should I do if presented with a subject that contains no highlights? If say there are no whites in the object should I shoot correct exposure manually using some nearby reference, or overexpose, using histogram as guide? The latter seems wrong to me but I would like guidance. Correct exposure would seem more logical. I am sorry if I am barking up the wrong tree here.
Peter
-
Hi,
from my point of view, it makes sense to use Expose to the right only if you shoot in RAW. Also, I would always recommend to bracket your exposure, so you always have a fall-back option in case highlight details are missing.
There are several good articles in the www that elaborate on the costs and benefit of that matter. You should draw your on conclusions based on some sample shots of your normal subjects.
-
If say there are no whites in the object should I shoot correct exposure manually using some nearby reference, or overexpose, using histogram as guide? The latter seems wrong to me but I would like guidance. Correct exposure would seem more logical. I am sorry if I am barking up the wrong tree here.
First off, ETTR is correct exposure for Raw data.
Next, just attempt to expose based on the scene, putting as much data as possible in that first stop of the tone scale that ETTR usually reserves for "highlights".
-
what should I do if presented with a subject that contains no highlights? If say there are no whites in the object should I shoot correct exposure manually using some nearby reference, or overexpose, using histogram as guide?
Correct me if I am wrong Peter, but reading your thoughts about the subject's highlights I have a feeling you are trying to understand ETTR from a film photography point of view, where camera exposure was linked to the desired final printed exposure.
If you want to maximise the quality (basically signal to noise ratio) of your captures shooting RAW, forget about that link. Capture exposure is not linked to the printed exposure anymore since in digital exposure can be adjusted without any loss or change in the image information.
No matter how the lights are in the scene, ETTR means exposing as much as possible so that the highest luminosity area of your scene gets the maximum RAW values right before clipping. Afterwards, in the RAW development process you will adjust exposure to your desire, being now certain that noise was minimised in your shadows.
BR
-
Hi,
I am sorry if this is a stupid question. But being new I would like some advice. I read a lot about exposure to the right and understand that not doing so underuses the capabilities of modern sensors. However, what should I do if presented with a subject that contains no highlights? If say there are no whites in the object should I shoot correct exposure manually using some nearby reference, or overexpose, using histogram as guide? The latter seems wrong to me but I would like guidance. Correct exposure would seem more logical. I am sorry if I am barking up the wrong tree here.
Peter
Actually, when there are no whites in the subject is when ETTR is most appropriate. If there are whites (and what I mean is near-whites that contain some detail, such as snow), then ETTR has the potential to blow them out so the detail is lost - although shooting RAW lessens the change of this happening.
Peter
-
Correct me if I am wrong Peter, but reading your thoughts about the subject's highlights I have a feeling you are trying to understand ETTR from a film photography point of view, where camera exposure was linked to the desired final printed exposure.
If you want to maximise the quality (basically signal to noise ratio) of your captures shooting RAW, forget about that link. Capture exposure is not linked to the printed exposure anymore since in digital exposure can be adjusted without any loss or change in the image information.
No matter how the lights are in the scene, ETTR means exposing as much as possible so that the highest luminosity area of your scene gets the maximum RAW values right before clipping. Afterwards, in the RAW development process you will adjust exposure to your desire, being now certain that noise was minimised in your shadows.
BR
Your explanation is very clear. I've been exposing to the right for years now, and it just occurred to me as I read your description that it should be fairly simple for the camera manufacturers to program their pro cameras' exposure systems to automatically put the brightest part of the scene right at the limit before clipping occurs, tailored for the specific model's sensor. While some people might object to more automation it certainly would not have to preclude the possibility of exposure compensation or purely manual exposure. It could even operate only when the camera is in RAW mode.
I for one would really like such a setup.
-
Your explanation is very clear. I've been exposing to the right for years now, and it just occurred to me as I read your description that it should be fairly simple for the camera manufacturers to program their pro cameras' exposure systems to automatically put the brightest part of the scene right at the limit before clipping occurs, tailored for the specific model's sensor. While some people might object to more automation it certainly would not have to preclude the possibility of exposure compensation or purely manual exposure. It could even operate only when the camera is in RAW mode.
Camera manufacturers PLEASE: when RAW histograms and an ETTR mode? (http://luminous-landscape.com/forum/index.php?showtopic=33267)
-
Camera manufacturers PLEASE: when RAW histograms and an ETTR mode? (http://luminous-landscape.com/forum/index.php?showtopic=33267)
It is ironic that given all the EEs working on the design of these things, such an obvious design optimization has never been included. Perhaps they tried too hard to mimic film cameras as a design spec, and there is nothing like that even possible in a film camera. In digital, you know everything there is to know about the "film" (sensor). You surely should leverage that information!
-
First off, ETTR is correct exposure for Raw data.
Next, just attempt to expose based on the scene, putting as much data as possible in that first stop of the tone scale that ETTR usually reserves for "highlights".
What exposure is correct is debatable, depending on the definition of correct. My understanding of ETTR is that the highlights containing important detail should be placed just short of clipping and no headroom is reserved for the highlights. A perfect ETTR exposure would have the highlights at but not above clipping, and IMHO, this would be the correct esposure. In practice, exposures not quite meeting this goal can be corrected with the exposure control and highlight recovery in the raw converter. With advances in sensor design with low noise and high full well electron capacity, overexposure is more harmful than under exposure, since clipping with overerxposure results in permanent loss of data. Shadow clipping can occur with underexposure, but shadow detail is usually limited by noise.
-
It is ironic that given all the EEs working on the design of these things, such an obvious design optimization has never been included. Perhaps they tried too hard to mimic film cameras as a design spec, and there is nothing like that even possible in a film camera. In digital, you know everything there is to know about the "film" (sensor). You surely should leverage that information!
I'm afraid that the reasoning is more like this: "Yes, we are selling a professional DSLR, for between $4000 and $8000. But there may be some rich beginners moving up from their $200 point-and-shoot cameras who will shoot jpegs and complain that they have 'blinkies' in them, worse than in their P&S camera." I think fear of losing such customers rather than the pros who are a "captive audience" might be behind the reluctance.
-
What exposure is correct is debatable, depending on the definition of correct. My understanding of ETTR is that the highlights containing important detail should be placed just short of clipping and no headroom is reserved for the highlights. A perfect ETTR exposure would have the highlights at but not above clipping, and IMHO, this would be the correct esposure.
I wouldn't debate that definition of correct although the bit about "no headroom reserved for highlights" I'm not clear on. I'd say you'd be reserving this assuming there's something in the highlight you wish to record (not clip). As image creators, we have to view the scene and decide what we hope to record. A shot of a black cat on coal is going to be treated differently than a white on white shot composed of chrome objects with highlight detail we may or may not wish to capture.
-
I wouldn't debate that definition of correct although the bit about "no headroom reserved for highlights" I'm not clear on. I'd say you'd be reserving this assuming there's something in the highlight you wish to record (not clip).
I am assuming that the highlights that need to be recorded are at or just short of clipping and anything above this would be allowed to clip. Leaving a half stop of "head room" is reasonable, but then you are not really exposing fully to the right, IMHO. However, if your intention is to leave a half stop of highlight headroom, then the exposure would be correct. One really needs to know how much headroom is reserved by the camera exposure system. If your camera metering allows a half stop of headroom and uses a hot TRC (such as many current Nikons), you might not want to add another half stop according to the histogram.
-
Correct me if I am wrong Peter, but reading your thoughts about the subject's highlights I have a feeling you are trying to understand ETTR from a film photography point of view, where camera exposure was linked to the desired final printed exposure.
If you want to maximise the quality (basically signal to noise ratio) of your captures shooting RAW, forget about that link. Capture exposure is not linked to the printed exposure anymore since in digital exposure can be adjusted without any loss or change in the image information.
No matter how the lights are in the scene, ETTR means exposing as much as possible so that the highest luminosity area of your scene gets the maximum RAW values right before clipping. Afterwards, in the RAW development process you will adjust exposure to your desire, being now certain that noise was minimised in your shadows.
BR
I think we are in agreement. What I meant is that a scene that does not have highlights can be exposed MORE to the right than a scene with highlights (again, I mean very light areas with detail) simply because you can increase the exposure more in the former without blowing out the lightest areas.
Peter
-
I often see the debate about "headroom" in conjunction with raw and ETTR. I don't understand the issue at all. ETTR is not reasonable without the photog knowing the effect (instead of judging it). This can be achieved only if the histograms and/or the clipping indication reflect the state of the raw data. If this is so, then any headroom is a waste of DR capacity. If the "raw exposure" is not reliably shown by the camera, then ETTR is simply lottery.
-
Shadow clipping can occur with underexposure, but shadow detail is usually limited by noise.
But exposing to the right will collect more photons in those shadows, and that's always a good thing as far as noise is concerned. Of course, if one wants really deep shadows... ;-)
-
Of course, if one wants really deep shadows... ;-)
Then clip them in post (in the converter). I agree with you, get less noise in shadows, even if you later decide to clip there, which is something many find an appealing color appearance.
-
Is flare (veiling or otherwise) ever a downside to ETTR? For example, +3 EC by using a slower shutter, printed down -3 linear EC in raw conversion.
-
But exposing to the right will collect more photons in those shadows, and that's always a good thing as far as noise is concerned. Of course, if one wants really deep shadows... ;-)
Increased exposure as a result of exposing to the right will linearly increase the pixel values of all parts of the image--from shadows to highlights. If the shadows are too bright as a result of increased exposure, then the midtones and highlights will also be too light. The remedy is to decrease exposure with the raw converter.
-
What I've found for real-world shooting, is that ETTR works best with lower-contrast scenes where you don't really have any whites. For scenes where there are highlight tones that I care about maintaining discernible detail, I get better results by placing them 1/3 to 1/2 stop below clipping, rather than right under the clipping limit. Even if that means having to add a small boost to the shadows or darker midtones, noise is pretty much a non-issue at base ISO and this approach gives me more natural-looking results with less fiddling around in the raw converter.
-
What I've found for real-world shooting, is that ETTR works best with lower-contrast scenes where you don't really have any whites. For scenes where there are highlight tones that I care about maintaining discernible detail, I get better results by placing them 1/3 to 1/2 stop below clipping, rather than right under the clipping limit. Even if that means having to add a small boost to the shadows or darker midtones, noise is pretty much a non-issue at base ISO and this approach gives me more natural-looking results with less fiddling around in the raw converter.
Since the original post by Michael on ETTR, it has become almost a religion for some photographers. However, the originally stated rationale for ETTR concerning the number of levels in the brightest f/stop of the image is incorrect and the naming of the process is based on the appearance of the the histogram, and the appearance of the histogram can be misleading. The basic concept of ETTR is to give as much exposure as practicable in order to collect the maximum number of photons, therey increasing the signal to noise ratio and maximizing dynamic range. An image taken at high ISO with the histogram fully to the right may have captured fewer photons than an image taken at a lower ISO and showing a histogram fully to the right. These principles are explained in great detail by Emil Martinec (http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/).
Some purists (such as Panopeeper) allow no headroom and place the highlights at clipping, while Jeff and others allow some headroom for protection of the highlights. If the sensor is linear up to clipping, headroom should not improve the image but can help prevent inadvertent clipping of the highlights. The advangtages of ETTR can be overblown. Since signal to noise varies as the square root of exposure, doubling of the exposure will improve S:N only by a factor of 1.4. With current sensors, this may not make much difference.
Maximizing the number of photons collected requires the use of base ISO, so the principles of ETTR can only be fully realized at base ISO. If shutter speed/aperture considerations prevent fully exposing to the right at base ISO, then one can use a higher ISO. Under these conditions, shadow S:N will be improved only to a certain point by increasing the ISO and having a histogram with data on the right. Above a certain ISO (often 1600 on many cameras), the histogram will look better with a higher ISO but the signal to noise ratio in the shadows will not be better than would be obtained at ISO 1600 with a histogram to the left. At this point, it is best to increase exposure in the raw converter. The higher ISO under these conditions will limit headroom and dynamic range. This is explained in detail in Emil's paper.
While these concepts are beyond the beginner lever, the take home message is to give as much exposure (shutter speed and f/stop) as possible. Signal to noise is largely deterined by exposure. A high ISO image will have more noise because of less exposure rather than the use of a high ISO per se.
-
Since signal to noise varies as the square root of exposure, doubling of the exposure will improve S:N only by a factor of 1.4. With current sensors, this may not make much difference.
You know Bill this applies only to photon noise, which is the least important in case of underexposure, i.e. in those cases where noise is really an issue.
Read noise, which is almost constant no matter the exposure, is the limiting factor in the deep shadows and hence determine DR. Here SNR is improved by a factor of 2 when doubling exposure.
Very deep shadows in the dark window of this high DR scene, allowing 1 stop headroom (left, indicated as +2EV) and extreme ETTR (right, indicated as +3EV):
(http://www.guillermoluijk.com/article/spot/ventana.jpg) . (http://www.guillermoluijk.com/article/spot/ruido.jpg)
RAW histograms:
(http://www.guillermoluijk.com/article/spot/ventana.gif)
The +2EV and +3EV indications are the exposure compensations applied over the spot metering on the white wall. The few clipped G pixels in the extreme ETTR had no visible effect at all in the final result.
PS: BTW this reminds me I owe you an email; I cannot think of any solution for the XP/Histogrammar issue, sorry.
-
You know Bill this applies only to photon noise, which is the least important in case of underexposure, i.e. in those cases where noise is really an issue.
Read noise, which is almost constant no matter the exposure, is the limiting factor in the deep shadows and hence determine DR. Here SNR is improved by a factor of 2 when doubling exposure.
Very deep shadows in the dark window of this high DR scene, allowing 1 stop headroom (left, indicated as +2EV) and extreme ETTR (right, indicated as +3EV):
An excellent point, Guillermo, and a very good ilustration. Read noise does predominate in the deep shadows, limiting dynamic range. However, current dSRL sensors have low read noise, so for most of the range of the sensor, shot noise predominates. This is shown in Figure 12 of Emil's paper (http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/noise-p2.html#read_vs_iso). The slope of the SNR plot is 0.5 down to about 6 stops below clipping, reflecting the effect of shot noise. With less exposure, the slope increases to unity, reflecting the effect of read noise. Like you, I am a proponent of ETTR, but one should not carry it too far. With my Nikon D200, underexposure is a killer, but with my D3, with much better read noise and a larger pixel size, maximum ETTR is less crucial.
-
Read noise, which is almost constant no matter the exposure, is the limiting factor in the deep shadows and hence determine DR. Here SNR is improved by a factor of 2 when doubling exposure
Guillermo, how did you come to this result? This is certainly not correct.
-
current dSRL sensors have low read noise, so for most of the range of the sensor, shot noise predominates. This is shown in Figure 12 of Emil's paper (http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/noise-p2.html#read_vs_iso). The slope of the SNR plot is 0.5 down to about 6 stops below clipping.
Correct Bill, but looking at the SNR values I wouldn't consider 6 stops below clipping the critical zone where to worry about noise. Even on my noisy 350D you can struggle and succeed to find texture 8 stops below clipping, so I prefer to go to lower RAW exposures to find the limit.
This is a Canon 20D plot done from real read noise and photon measures from Emil. When SNR reaches 2EV (12dB is a good photographic criteria to still be able to recognize textures), we are already more than 8 stops far from clipping and in that zone the slope is clearly almost 1.0:
(6dB/EV = slope 1.0)
If you look at Emil measures, they are always parallel lines (http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/noise1d3.gif) in the areas before and after the toe.
Regards.
-
Guillermo,
I find your effort of turning the curves of Gabor Schorr (who is that?) in straight lines pretty useless. Though I agree with your I wonder why.
Why don't you try it more accurately, taking specific numbers measured on the raw data?
Example (the noise is expressed in raw pixel standard deviation, i.e. it is an absolute value), at the indicated intensity, measured from clipping:
Canon 40D, ISO 100
- EV StDev
6.0 11.0
8.0 7.6
7.0 8.4
9.0 6.8
Increasing the amount of light by two stops results in increasing the noise by 45% respectively 24% - this in contrast to quadrupling the signal.
You can use the raw image Marc McCalmont created for measurement, http://www.panopeeper.com/Download/Canon5D...O00100_1221.CR2 (http://www.panopeeper.com/Download/Canon5DMkII_CCC_ISO00100_1221.CR2) (5D2, ISO 100), as an example.
-
Gabor, the data came from your Excel file! they were numbers, not curves.
-
Gabor, the data came from your Excel file! they were numbers, not curves.
Gullermo,
the graphs I presented don't go over 80% noise (SRN=1.25), and already that is too far. There is a problem in principle with the measurement of the noise: as soon as black clipping occurs, i.e. some of the pixel values are zero - or under zero in Canon files - the black level corrected standard deviation is lower than the real one. The clipping starts at about -9.5 EV with the 5D2, ISO 100.
Example from the 5D2 ISO 100 file: at -11 EV, the standard deviation is 6.3 absolute, but only 5.6 after BL correction.
Thus the numbers at the dark end are useless. I did not "publish" the uncorrected numbers, because they are not comparable to other cameras.
Now, specific examples from the 5D2 ISO 100 sample:
-EV StDev Noise %
9.27 6.74 28.3
8.27 7.11 14.9
7.23 8.06 8.23
The step from -9.27 EV to -8.27 EV in fact halves the noise ratio, as you stated - but from -8.27 to -7.23 the reduction is less.
The limit of clipping at ISO 1600 is about -8 EV. Samples not darker than this limit show
-EV StDev Noise %
8.07 17.8 32.6
7.09 22.3 20.8
5.97 13.0 30.5
this is far from being doubled by 1 stop difference in the exposure. Note, that the amount of captured light here is 1/16 of that that with ISO 100 with the same pixel value.
I don't know the reason of the difference from the "ideal" result; I know nothing of the hardware, but I think the "read noise" has several sources and I guess not all are constant.
-
I don't know the reason of the difference from the "ideal" result; I know nothing of the hardware, but I think the "read noise" has several sources and I guess not all are constant.
Seems so, here's one explanation why
http://astrosurf.com/buil/5d/test.htm (http://astrosurf.com/buil/5d/test.htm)
And some more data here
http://astrosurf.com/buil/50d/test.htm (http://astrosurf.com/buil/50d/test.htm)
A few points worth noting imho
Buil's protocols are the correct ones, from the point of view of maximizing SNR, but hardly practical for standard photography (who wants to substract master biases made of 20+ frames from a shoot of a building at night).
Maximizing SNR involves, at the most basic level, minimizing noise whatever the origin is, and getting as many photons as one can for a given exposure. In other words, this is exposing as far "right" as one can without overexposing and staying in the sensor linear response zone if photometric measures are to be made, probably of little practical interest for photography... except when testing to the limit for the sake of it.
The behaviour of cameras changes somewhat as the sensor heats (dramatically as far as the thermal noise component is concerned). It may seem again like a non-issue for photography, but if one does extensive testing consisting of almost continuously shooting, live view usage, etc, the sensor will be warmer after a while if there's no cooldown period between shots and basic camera use.
Camera manufacturers use lots of tricks to minimize noise, this is why Nikon cameras, with their "artificially" low noise are not suitable, or at least the best choice, for astrophotography. The Nikon strategy could very well be the best for everyday photography, but it also means that the ideal "ETTR" strategy will vary with the camera model just as the ideal ISO, as far as maximizing SNR is concerned, varies between cameras of the same brand.
And beyond deep pixel peeping, the factor that matters most, in my observations and for my purpose, is heat. The difference in image quality between a warm sunny afternoon and a colder morning is striking. Keeping the camera in cold storage, if practical, beats ETTR by a wide margin. But of course, one can try to do both.
-
Seems so, here's one explanation why
I don't underdstand French. As much as I can interpret that paper, it does not explain why the noise classified as "read noise" is not constant within a single shot.
And some more data here
Again, no explanation. However, honestly if someone would explain the hardware reason, I could only repeat it anyway.
On the other hand, I have an observation re one of the points, "quantum efficiency": it can not be measured without removing the color filters; and if that could be accomplished at all, then the microlenses too would be eliminated, making the result irrelevant.
Another observation: beside heat, the individual camera copy counts a lot. If I were interested on the cleanness to that degree as astrophotographers are, I would make raw shots from several cameras under identical circumstances and pick the one with the lowest read noise (measured on the masked pixels).
-
On the other hand, I have an observation re one of the points, "quantum efficiency": it can not be measured without removing the color filters; and if that could be accomplished at all, then the microlenses too would be eliminated, making the result irrelevant.
QE can definitely be measured with the bayer matrix. Why couldn't it be? All you need to do is to count incoming photons, either from calibrated sources (or known sources, see the link) and then count the number of electrons. For example
http://www.clarkvision.com/imagedetail/dig...d.qe/index.html (http://www.clarkvision.com/imagedetail/digital.photons.and.qe/index.html)
and more info here
http://www.clarkvision.com/imagedetail/dig...rmance.summary/ (http://www.clarkvision.com/imagedetail/digital.sensor.performance.summary/)
Some manufacturers, for example Kodak, provide QE numbers for
http://www.kodak.com/global/plugins/acroba...022LongSpec.pdf (http://www.kodak.com/global/plugins/acrobat/en/business/ISS/datasheet/interline/KAI-04022LongSpec.pdf)
you will see (page 20) that this QE is "measured"
There is nothing special to a Bayer Matrix that would prevent the comparison the amount of incoming photons to the amount of generated electrons. There is even the notion of Geometric QE which, for a given surface area takes into acount that 50% of that area has a specific G QE, 25% R QE and 25% B QE
As far as the read noise variation with ISO is concerned, the Martinec paper addresses the issue in English.
-
QE can definitely be measured with the bayer matrix. Why couldn't it be?
It has nothing to do with the Bayer arrangement but with the microfilter and with the color filter. How on earth do you calculate with the number of photons, when you don't know the proportion passing through the filter and directed in the well?
All you need to do is to count incoming photons, either from calibrated sources (or known sources, see the link) and then count the number of electrons. For example
I find Clark's calculations childish.
Some manufacturers, for example Kodak, provide QE numbers
I know and I regard that as QE.
As far as the read noise variation with ISO is concerned, the Martinec paper addresses the issue in English
I am looking for an explanation of the variation in read noise at a given ISO; even more, in a single shot.
-
I would like to thank everyone who replied to my OP, especially GLuijk whose early reply answered my question. One further question though. I can see 2 possible ways of increasing the exposure. I use a Canon 5D and 350D.
1. Shoot Aperture or Shutter priority with exposure compensation set +2. Spot metering from lightest area of subject. This semms to me to limit the EV to +2. Lock exposure and focus and shhot.
2. Shoot Manual with spot metering and then dial in reqired EV. This does not limit one to +2.
Is there apreferred way of doing it used by you all?
Peter
PS The replies by the end became very technical and left me dizzy.LOl
-
I would like to thank everyone who replied to my OP, especially GLuijk whose early reply answered my question. One further question though. I can see 2 possible ways of increasing the exposure. I use a Canon 5D and 350D.
1. Shoot Aperture or Shutter priority with exposure compensation set +2. Spot metering from lightest area of subject. This semms to me to limit the EV to +2. Lock exposure and focus and shhot.
2. Shoot Manual with spot metering and then dial in reqired EV. This does not limit one to +2.
Is there apreferred way of doing it used by you all?
Peter
PS The replies by the end became very technical and left me dizzy.LOl
To use ETTR effectively, you have to know the headroom that the camera allows for the highlights. For example, if you expose so that the highlights are just short of clipping according to the histogram and camera allows 0.5 EV of headroom for the highlights, the highlights in the raw file will be 0.5 EV short of clipping and you have not exposed fully to the right. You have to conduct tests using the camera JPEG processor or the raw converter and compare the results to those obtained by a program which shows the raw data (Iris, DCRaw, etc). See this article on Libraw (http://www.libraw.org/articles/Canon-5Dmk2-headroom.htmlhttp://www.libraw.org/articles/Canon-5Dmk2-headroom.html) for an example.
To implement ETTR, the purist would use spot metering to take a reading from the highlights, use manual exposure, and then give 2-3 EV additional exposure (depending on previous tests) to bring the highlights up to the desired value just short of clipping. A common value for the exposure increment is +2.5 EV. If you use aperture priority or shutter priority, you would judge exposure by the histogram (after performing the above mentioned tests) and use exposure compensation to set the highlights. In this case, a fixed compensation is not appropriate. For example, if you are shooting snow scenes you would have to use a larger value than for a normal scene, since the camera will render the snow as mid gray. If you use evaluative exposure (matrix with Nikon), the camera has already added a compensation and the compensation you dial in will be added to the automatic compensation.
-
> It has nothing to do with the Bayer arrangement but with the microfilter and with the color filter.
On one side you have incoming photons, on the other side you have outgoing electrons.
> How on earth do you calculate with the number of photons,
> when you don't know the proportion passing through the filter and directed in the well?
Well, with a calibrated source, eventually as a star, you merely take into account the number of photons you measure vs the ones that arrive. There are a few ways to do that, but it won't be necessary to get into details...
> I know and I regard that as QE.
Great! Then you have just discovered the solution to your question above.
Take an arbitrary source.
Take shot(s) with that Kodak chip. Get rid of noise (through standard calibration).
Take shot(s) with the camera you want to measure. Get rid of noise as above.
Compare the two recorded signals.
That's it, there you go, you have relative QE.
> I am looking for an explanation of the variation in read noise at a given ISO; even more, in a single shot.
OK - that's an easy one too.
There are two components to read noise
- noise during analog to digital conversion: this conversion is not perfectly repeatable. There is a statistical uncertainty, not necessarily gaussian. (See Merline - Howell, 1996)
- noise introduced by the electronic chain itself (for example the size of the amplifier, its sensitivity and, most importantly, its temperature). In a CCD, simply reading a line will raise the temperature of the amplifier. In a CMOS sensor, the problem is a bit different because each pixel sensing well has its own amplifier. But this introduces another source of non uniform response as those amplifiers do not behave identically. In a way, you have hot and cold amplifiers just as you have hot and cold pixels. It also introduces leakage currents. Etc... etc... BTW, you can't read/convert/work as fast as you want, because this does increase heat significantly, which is why there are now multiple channel electronics.
The killer factor is always heat.
-
2. Shoot Manual with spot metering and then dial in reqired EV. This does not limit one to +2.
I strongly recommend shooting in M mode for ETTR, where you can overexpose as much as you want.
I have the same cameras as you (350D and 5D). Bill (bjanes) explains clearly the way to do ETTR; to find out more about how much headroom these cameras allow from light metering to saturation have a look at this article (ignore the Spanish text if you don't understand, the sample images can easily be interpreted, MEDICIÓN PUNTUAL=SPOT METERING): ETTR WITH SPOT METERING (http://www.guillermoluijk.com/article/spot/index.htm). You can use spot metering on the 5D which allows up to +3EV over spot metering if only the highlights entered the metering circle. Some +2.5EV as Bill suggests is probably the best tradeoff between quality and safety.
The problem to chek in place if your ETTR succeded, is that camera displays are not RAW-oriented but JPEG-oriented, which is an overexposed processing of the RAW file, i.e. the camera will report pesimistic information about how you exposed. To learn more on how to make your camera display (histograms and highlight clipping) closer to the real RAW condition: UNIWB. MAKE CAMERA DISPLAY RELIABLE (http://www.guillermoluijk.com/tutorial/uniwb/index_en.htm). You will find there files to implement UniWB on both your cameras.
Taking pictures was never so easy as with digital photography... they said.
Regards.
-
Taking pictures was never so easy as with digital photography... they said.
I sense here a certain sarcasm, Guillermo. However, I believe it really is true. Taking pictures has never been so easy. I paid A$850 for my first 1GB Micro drive, for my Canon D60. One can now buy 32GB compact flash cards for very much less.
Modern DSLRs now takes 5 or more frames per second. If anyone has some doubt about his ability to get a good ETTR exposure, then bracket all shots. Problem solved.
-
Modern DSLRs now takes 5 or more frames per second. If anyone has some doubt about his ability to get a good ETTR exposure, then bracket all shots. Problem solved.
My crappy Canon DSLR supports only three shots in a bracket; this is a shame, particularly because the shots can be made in a small fraction of a second with MLU and live view combined. A reasonable camera software could even automatically delete the files of those shots from the bracket, which are definitively worthless compared to the others.
-
Hi Gullermo,
Reading this thread with interest and have started to experiment with the UNIWB method.
But I have a question regards WB and White Point estimation.
If for example I ETTR and say clip all channels in the RAW file ( too much ETTR), how will that effect the WB.
For example I cannot go back and reshoot the scene.
what are the main methods used for white balancing , ie Gray world, Retinex, Colour by Convolution etc or are the channels balanced to a set white point ( daylight, tungesten etc)
When I open a RAW in ACR the white balance 'as shot' is displayed. how is calculated? And how would it be affected if the RAW is overexposed as mentioned above.
Im also working in DCRAW and Matlab.
-
I strongly recommend shooting in M mode for ETTR, where you can overexpose as much as you want.
I have the same cameras as you (350D and 5D). Bill (bjanes) explains clearly the way to do ETTR; to find out more about how much headroom these cameras allow from light metering to saturation have a look at this article (ignore the Spanish text if you don't understand, the sample images can easily be interpreted, MEDICIÓN PUNTUAL=SPOT METERING): ETTR WITH SPOT METERING (http://www.guillermoluijk.com/article/spot/index.htm). You can use spot metering on the 5D which allows up to +3EV over spot metering if only the highlights entered the metering circle. Some +2.5EV as Bill suggests is probably the best tradeoff between quality and safety.\
Guillermo,
A few points for discussion. I don't understand Spanish, but the illustrations are more or less self explanatory. The histograms shown by Histogrammar are excellent, but the program does not work with my current XP machine due to problems in the graphics driver. As far as I know, one still needs to use DCRaw or something something similar to process the raw file. Personally, I prefer Iris, since it has a graphical interface. Another option is to use Rawnalize (Gabor), which can show the raw histogram directly from the raw file and has a graphical interface.
The problem to chek in place if your ETTR succeded, is that camera displays are not RAW-oriented but JPEG-oriented, which is an overexposed processing of the RAW file, i.e. the camera will report pesimistic information about how you exposed. To learn more on how to make your camera display (histograms and highlight clipping) closer to the real RAW condition: UNIWB. MAKE CAMERA DISPLAY RELIABLE (http://www.guillermoluijk.com/tutorial/uniwb/index_en.htm). You will find there files to implement UniWB on both your cameras.
Since the red and blue multipliers are greater than unity for daylight and most other illuminatioin, one can use the RGB histograms on most of the more adavnced cameras to check for clipping. If the green channel is truly clipped, one must reduce exposure. If the scene contains strong reds or blues and clipping in these channels is present, this could be due to the multipler greater than unity or true clipping in the sensor. One should then use UNIWB. I keep normal white blance in one custom profile on the camera and UNIWB in another.
Saturation clipping can also occur if the camera is rendering into a small color space such as sRGB. I always set my camera to aRGB, which is the widest space on Nikon cameras. Does Canon offer a wider space for JPEG? A low saturation setting on the camera can reduce saturation clipping.
If the camera allows considerable headroom and applies an S curve to the data, a strong S curve with a high contrast setting can also cause clipping in the histogram. Many photographers set the camera to low contrast.
Regards,
Bill
-
In addition to all that, another thing I find bad in UniWB is that output profiling means a weighted mix of all three channels data. So if the output G is a function of R, G and B, how much can we trust camera clipping? if output G is not clipped, is because it is really not clipped in the RAW or because the influence of R and B made it appear as non-clipped after the matrix linear combinations done for camera colour profile conversions?
Anyway I like to use it, not because of its accuracy but because it's more stable (you know better what to expect, as opposite to selecting a particular WB). BTW in my cameras multipliers are always >=1 so the R and B channels can clip due to WB (in fact with ETTR'ed RAW files they usually do if UniWB is not set).
Regards.
-
If for example I ETTR and say clip all channels in the RAW file ( too much ETTR), how will that effect the WB.
what are the main methods used for white balancing , ie Gray world, Retinex, Colour by Convolution etc or are the channels balanced to a set white point ( daylight, tungesten etc)
When I open a RAW in ACR the white balance 'as shot' is displayed. how is calculated? And how would it be affected if the RAW is overexposed as mentioned above.
WB is independent of how much you exposed (clipped) the RAW channels. It just affects how the JPEG thumbnail displays on your camera, and how WB is applied when developing the RAW file in case you choose 'as shot'. The 'as shot' WB factors will be those you chose in the camera, whatever they were (auto, preset or custom WB).
For white balancing a clipped RAW file the same methods apply as with any other RAW file (auto, any preset, eyedropper over a neutral area of the scene,...). You simply won't be allowed to use the eyedropper (is this the name in English?) on the blown areas.
-
Since the red and blue multipliers are less than unity for daylight and most other illuminatioin, one can use the RGB histograms on most of the more adavnced cameras to check for clipping.
With which cameras? I don't think this is true for any of the Nikon's I've shot with over the last several years.
-
papa v2.0, the WB methods you discuss are automatic estimation methods. They are only relevant if you wish to understand how a raw conversion software performs its "automatic" WB mode. Most of the time, however, I suspect you are simply going to click-WB or choose one of the standard presets and then tweak from there. Ultimately you are just shoving the native pixel coordinate values through a 3x3 matrix. Simple white balance without color optimization just means multiplying each of the R, G, and B values by a separate constant -- i.e., the matrix is a diagonal matrix. You can refer to Chapter 6 of the DNG specification if you wish to learn the specific math that Camera Raw and Lightroom use.
-
papa v2.0, the WB methods you discuss are automatic estimation methods. They are only relevant if you wish to understand how a raw conversion software performs its "automatic" WB mode.
I am interested to find out what methods are use for calculating the auto WB, either in ACR or even by camera manufactures, although I suspect that they are proprietary and there is little information available. If anyone has any ideas...
-
With which cameras? I don't think this is true for any of the Nikon's I've shot with over the last several years.
Less than unity was a typo and I should have said more than unity. Obviously, if the green channel is intact (multiplier = 1.0), the red and blue channels could clip only with multipliers greater than 1.0.
-
It is ironic that given all the EEs working on the design of these things, such an obvious design optimization has never been included. Perhaps they tried too hard to mimic film cameras as a design spec, and there is nothing like that even possible in a film camera. In digital, you know everything there is to know about the "film" (sensor). You surely should leverage that information!
They have been trying super hard to make DSLR behave like film cameras, and now you are telling them that the DSLR should behave in a digital specific way?
Cheers,
Bernard
-
They have been trying super hard to make DSLR behave like film cameras...
Which is kind of silly considering they are not. The underlying capture process is as we all know, vastly different. So I can only assume they do this to comfort the masses to make more sales, not necessarily targeting the professional or those hopping for the bet capture data. But then if they all cared for our well being, we wouldn’t be dealing with the messiness of all these proprietary files every time a new camera comes out.
-
I am interested to find out what methods are use for calculating the auto WB, either in ACR or even by camera manufactures, although I suspect that they are proprietary and there is little information available. If anyone has any ideas...
I think there is not much secret to tell. Typical Auto WB tries to get an image which is gray in average, i.e. where the R, G and B averaged values are the same. And to achieve that it calculates the 3 appropiate relative coefficients (2 in practice) by which each RAW channel must be scaled.
Auto WB will never be able to achieve a correct WB in all situations because it ignores the photographer's intentions, which cannot be guessed by a piece of software. That is why sometimes it works fine and sometimes horrible. The success depends on the scene and user's perception rather than on the calculation method.
This scene was developed under 4 different WB coefficients. See how Zero Noise Auto WB nearly provides the same result as DCRAW Auto WB. The slight changes are due to subtle differences in the algorithms, but the goal was the same and to achieve it is very simple. The camera (Canon 350D) Daylight preset was too bluish for my taste. The last sample uses a round patch over the USA flag for WB, although I prefer the Auto WB in this case.
(http://www.guillermoluijk.com/misc/wb1.jpg)
(http://www.guillermoluijk.com/misc/wb2.gif)
Regards.
-
I think there is not much secret to tell. Typical Auto WB tries to get an image which is gray in average, i.e. where the R, G and B averaged values are the same. And to achieve that it calculates the 3 appropiate relative coefficients (2 in practice) by which each RAW channel must be scaled.
Regards.
As per Eric Chan's (http://luminous-landscape.com/forum/index.php?s=&showtopic=37042&view=findpost&p=305850) previous post, multiplying by the red and blue coefficients merely gives a first order approximation to the white balance (diagonal matrix). To obtain a more accurate white balance, one has to use a full matrix derived from the correlated color temperature of the set white balance. ACR interpolates between daylight and tungsten profiles to derive the matrix. The precise methodology used by ACR is described in Chapter 6 of the DNG specification, but many of these matters are beyond my expertise. For my Nikon D3, I don't know exactly what white balance information is recorded in the raw file. It could by multipliers, a correlated color temperature, or a CIE xy coordinate. Perhaps Eric can elaborate.
If you have a well defined white as in Guillermo's image, one can achieve an approximate WB relatively simply by using linear curves for the red and blue so as to equalize the RGB values in the whites and neutral areas. This is equivalent to using multipliers. This is shown in the following image rendered by Iris into a gamma one space without white balance. The white balance curves are shown. Additional curves were needed for gamma, contrast, and saturation.
[attachment=16247:WhiBal.png]
The image is hardly optimized, as shown by comparison to the ACR rendering.
[attachment=16248:2009Jul16_0014.jpg]
-
And I thought this was the beginners section I'd hate to read the advanced section!
Marc