Luminous Landscape Forum
Raw & Post Processing, Printing => Adobe Lightroom Q&A => Topic started by: Robert Boire on December 30, 2011, 12:08:27 am
-
Hi,
Some of the posts in a recent thread on ETTR (http://www.luminous-landscape.com/forum/index.php?topic=60049.0) got me to puzzling over how the Exposure slider works in general in LR. The general idea being that ETTR increases the amount of light received at the sensor and that the "over-exposure" that occurs (assuming the highlights are not blown) can be corrected in post by decreasing the exposure with the exposure slider.
Now if each adjustment of +/- 1 Ev in post is equivalent to +/- one stop on the camera, I would have thought that adjusting the exposure slider should simply shift the entire histogram left or right - ie shift the values from one level to another without changing the relative values of the peaks themselves. This is clearly not what happens. Adjusting the exposure changes not only the position but the shape of the histogram as well - even when there is room at the high and low ends so that clipping is not an issue.
What am I missing?
And it the exposure slider is adjusting...exposure.. then what is the brightness slider adjusting?
Thanks
-
According to "Real World Camera Raw with Adobe Photoshop CS5", the brightness slider basically "lets you redistribute the midtone values without clipping the highlights or shadows."
I highly recommend purchasing this book for the version of Photoshop that you are working with. I have purchased one each time I purchased Photoshop beginning with CS. It is cornucopia of information for Camera Raw users.
-
What am I missing?
The fact that the histogram is gamma encoded maybe, with a sRG gamma that differs from what I'd like to think as the log2 nature of an exposure shift?
-
A log histogram (i.e. an histogram in EV divisions) actually shifts left/right with exposure. These are real log histograms from RAW files obtained at 1EV intervals:
(http://www.guillermoluijk.com/misc/sequence.gif)
Unfortunately LR, ACR or any camera's histograms are not logarithmic, so you can simply not expect the previous behaviour from them. It's just a matter of how the X-axis scale is chosen, or in other words a matter on the low degree of intelligence of the developers for not including such a trivial option in their software (this excludes our forum member Eric Chan).
Regards
-
Thanks Guillermo, very interesting.
So presumably shifting the exposure slider is equivalent to adjusting f-stop (speed, ISO) in the camera itself.
A question though..
I noticed that red and green separate as the EV is decrease.. though their relationship to each other and the blue stay about the same. Why?
Also what are the patches on the right?
Would you mind telling me how you produced this histogram?
-
So presumably shifting the exposure slider is equivalent to adjusting f-stop (speed, ISO) in the camera itself.
A question though..
I noticed that red and green separate as the EV is decrease.. though their relationship to each other and the blue stay about the same. Why?
Also what are the patches on the right?
Would you mind telling me how you produced this histogram?
Yes, thanks to sensor linearity, the only difference between changing exposure in the RAW developer and in the camera is SNR (DOF and motion blur considerations aside). Pushing exposure through aperture/shutter, or even through ISO improves SNR, while pushing exposure in software doesn't.
These histograms are RAW histograms obtained from a series of shots against a white wall, that is why they are so narrow (very low dynamic range scene). The G channel is the winner because it usually is. Green photosites receive more light for two reasons: in real life there is more power in that area of the spectrum, and camera colour filters are weaker for the G channel. The patches on the right indicate the colour of the image without white balance applied, that is why it it greenish.
This would be a more typical log histogram, and the effect of altering exposure in the camera:
(http://www.guillermoluijk.com/misc/histogram.gif)
It corresponds to this image:
(http://www.guillermoluijk.com/misc/chica.gif)
Regards
-
RAW histograms obtained from a series of shots against a white wall
Now I am not sure I got it...
Did you change the exposure in camera or is it a single shot with exposure changed in post processing?
I tried a crude version of my own by taking a shot against a wall and then changing exposure in LR. The colors never separated. Would this be because LR does not have a log axis?
Regards
-
Now I am not sure I got it...
Did you change the exposure in camera or is it a single shot with exposure changed in post processing?
I tried a crude version of my own by taking a shot against a wall and then changing exposure in LR. The colors never separated. Would this be because LR does not have a log axis?
Sorry, I didn't give you enough information. I did a series of shots, changing exposure (shutter). The RAW files obtained were developed in a very special way, which can be considered RAW in terms of calculating the RAW log histogram:
- No white balance applied (-r 1 1 1 1): this means each RGB channel keeps its captured relative exposure.
- No output colour profile conversion (-o 0): this means the RGB channels do not get mixed on each other (in a regular RAW development, each of the three R'G'B' output channels is a function of all three RGB captured input channels).
To do this forget about using any commercial RAW developer, since they do not allow any of the two things above, all of them apply a white balance and convert to some output colour profile. I did it with DCRAW (http://www.guillermoluijk.com/tutorial/dcraw/index_en.htm). The command to do such thing is:
dcraw -v -r 1 1 1 1 -o 0 -4 -T file.cr2
The resulting TIFF's were input into Histogrammar (http://www.guillermoluijk.com/tutorial/histogrammar/index_en.htm) to obtain the log histograms.
Regards.
-
Green photosites receive more light for two reasons: in real life there is more power in that area of the spectrum, and camera colour filters are weaker for the G channel.
Not unless the laws of physics have changed since I took light and optics in college (admittedly a lot of years ago now). The shorter the wavelength of the light, the more energy it has, so you decrease the light's energy as you move from violet at the short end of the spectrum to red at the long end of the spectrum. Green is pretty much right in the middle (550nm). Maybe you are using 'power' in a different way than I think about it.
Alan
-
The light spectrum depends on the scene, and the max peak could not be around 550nm. But the combination of the sensor colour filters plus the entire light spectrum shape usually ends in the G channel accounting more photons than the R, and the R channel more than the B. Even in shots of a blue sky the G channel gets higher exposure than the B channel in most cameras and situations.
This is the RAW histogram:
(http://www.guillermoluijk.com/article/ettr/histlin.gif)
from this scene:
(http://www.guillermoluijk.com/article/ettr/final.jpg)
The image has strictly blues and reds, but G got higher exposure than R and B.
Regards
-
What you are seeing in the example you post is spectral overlap which is well documented. It has nothing to do with "power" of green (unless of course you are taking a picture of Green Lantern). I'm not trying to be difficult on this just making the point that it is not related to the energy of the photon which is inversely proportional to the wavelength of the light (UV stronger than infrared).
Edit Added - see THIS (http://www.gamutvision.com/docs/camera_scanner.html) for the spectral overlap of the Leica M-8 sensor
-
Alan, forget about the word 'power' or 'energy', it's just photon count. Just wanted to point that the G channel is the one that accounts for more photons in most real world scenes (producing higher RAW encoded levels), and this is after the combination of the spectral profile of light in usual situations plus the colour filters of the Bayer sensor.
Regards
-
The recorded intensity of the green channel may be higher, but I do not think this answers the original question. Which was why did the red and green channels separate from the blue in the first set of histograms posted by Guillermo? Has this got something to do with turning off the WB?
BTW thanks for the details?
R
-
To do this forget about using any commercial RAW developer, since they do not allow any of the two things above, all of them apply a white balance and convert to some output colour profile. I
iridient raw developer is a commercial and does (it has an option to set wb using per channel multipliers, albeit does not allow different G1 and G2 multipliers, and has an option to disable conversion to some output color profile).
-
Alan, forget about the word 'power' or 'energy', it's just photon count. Just wanted to point that the G channel is the one that accounts for more photons in most real world scenes (producing higher RAW encoded levels), and this is after the combination of the spectral profile of light in usual situations plus the colour filters of the Bayer sensor.
Regards
On that we can agree! :D Thanks for the continuing interesting posts on all of this. It would be great if camera makers could simplify everything and just record raw 'real' histograms in the camera. Maximizing SNR will be easier to accomplish.
-
why did the red and green channels separate from the blue in the first set of histograms posted by Guillermo? Has this got something to do with turning off the WB?
It's simply because the B channel is the one that collected less photons in my home's wall. In those histograms you see what the sensor saw, and the B photosites received less photons. After white balance the 3 channels would get perfectly aligned but I deliberately didn't apply WB.
This is a good example to explain why the B channel is said to be noisier than the other two. Actually it is not, what happens in most real life situations (like in my home's wall, under tungsten lighting) is that the B channel records much lower exposure than G and R, making its SNR worse (because given an ISO, SNR only depends on RAW exposure). But at the same level of exposure, all three channels have symmetrical SNR.
A way to obtain UniWB (http://www.guillermoluijk.com/tutorial/uniwb/index.htm) (WB cancellation) for a digital camera, is to find which precise colour produces the same RAW exposure on all three channels. It should be a colour similar to this:
(http://www.guillermoluijk.com/tutorial/uniwb/cartamagenta.jpg)
Shooting at that colour without applying WB on my Canon 350D, I obtained this RAW histogram (nearly perfectly aligned channels):
(http://www.guillermoluijk.com/tutorial/uniwb/cartamagenta_his.gif)
Regards
-
Thanks. All very interesting. I will need to spend some time on the applications you provided links to.
-
The resulting TIFF's were input into Histogrammar (http://www.guillermoluijk.com/tutorial/histogrammar/index_en.htm) to obtain the log histograms.
Guillermo,
I used Histogrammar ver 1.2 yesterday and it worked fine, but this morning I got a message that the program had expired and I should download a new copy. I downloaded the latest copy and got the same message. Please advise.
I am posting this message publicly since others are undoubtedly interested in using the program. Those who do find the program useful should make a donation (I have). Thanks to Guillermo for his contributions to this forum and wishes for a happy and productive new year.
Bill
-
OK, I'll upload a new build ASAP. The old one will work if you set any date before 2012.
I thought I had already sent you a never-expiring copy. I will.
-
OK, I'll upload a new build ASAP. The old one will work if you set any date before 2012.
I thought I had already sent you a never-expiring copy. I will.
Thanks,
Bill
-
See Bryan's comment, below, which is 100% correct. There seems to be a lot of confusion over what the exposure slider is supposed to do, probably because Adobe chose a really poor name for it. The exposure slider is not, and isn't supposed to be, the same as adjusting exposure in camera. It doesn't shift the all portions of the image up or down equally as adjusting the f-stop does (assuming there's no clipping of the image).
Hover your mouse pointer over the exposure slider while watching the histogram. Notice that part of the histogram background turns a lighter gray? THAT'S the portion of the image primarily affected by the exposure slider.
Now drag the Recovery slider to the right, and hover the mouse over that slider. Notice that a different part of the background turns a lighter gray? Do the same thing with the Fill and Blacks sliders. Each slider primarily affects a different part of the histogram, and the image.
According to "Real World Camera Raw with Adobe Photoshop CS5", the brightness slider basically "lets you redistribute the midtone values without clipping the highlights or shadows."
I highly recommend purchasing this book for the version of Photoshop that you are working with. I have purchased one each time I purchased Photoshop beginning with CS. It is cornucopia of information for Camera Raw users.
-
I can not see the histogram (or the histogram background) change at all when I hover my mouse over any of the sliders in ACR 6.6 or ACR 4.6 (Windows 7 64 bit). I tried this both before and after moving the sliders.
-
That only works in Lightroom...
-
There seems to be a lot of confusion over what the exposure slider is supposed to do, probably because Adobe chose a really poor name for it. The exposure slider is not, and isn't supposed to be, the same as adjusting exposure in camera. It doesn't shift the all portions of the image up or down equally as adjusting the f-stop does
There is no confusion, and the name is not poorly chosen. The exposure slider really does have the effect of "changing exposure" or basically linearly multiplying the channels by a factor. (It would not be possible to do ETTR if this was not true).
The best explanation I have come across is in this thread (http://www.luminous-landscape.com/forum/index.php?topic=36161.msg296848#msg296848) , where Guillermo Luijk shows the difference between exposure and brightness, and a few posts later, Eric Chan explains that when there are blown out areas, the exposure slider will invoke automatically highlight recovery
-
Duh.....I guess that is the reason why the thread is in the Adobe Lightroom Q&A...LOL. I did not pay attention to that before I made my statement. Ha! Thanks Jeff.
I don't know that I would ever use the feature if it were in ACR....but...it might make us ACR'ers feel like we are as important (smart) as the Lightroomers. We like to play with the toys too! ::) Hint, hint Adobe.
-
The Exposure slider in ACR and Lightroom is intentionally measured in stops (e.g., from -4 to +4 stops), and for the most part (*) it does correspond to the standard notion of photographic exposure. For example, if you set Exposure to +1, this will double the exposure of the image. The reason it may seem as if other non-linear stuff is going on, is because images are rendered in ACR/LR with tone curves -- e.g., for shaping the highlights and shadows. For example, the default point curve of Medium Contrast is part of this tone shaping. If you adjust exposure, then this affects how the values get mapped to the tone curve.
While a log-base-2 histogram is useful for determining exposure levels, it is not particularly useful for judging rendered output for a color image once tone mapping is applied. Remember, the histogram in ACR/LR reflects the rendered output (output-referred image), not the original input image (scene-referred input).
Eric
(*) The reason I say "for the most part" is because of course there are limits. For example, in the field if you were to reduce the exposure (e.g., by halving the exposure time), you can record more and more highlight detail. But if you try to do this in a raw conversion software, of course that doesn't work. In other words, reducing the Exposure slider can reveal a bit more highlight detail, but very soon you exhaust what was actually recorded in the image.
-
The Exposure slider in ACR and Lightroom is intentionally measured in stops (e.g., from -4 to +4 stops), and for the most part (*) it does correspond to the standard notion of photographic exposure. For example, if you set Exposure to +1, this will double the exposure of the image. The reason it may seem as if other non-linear stuff is going on, is because images are rendered in ACR/LR with tone curves -- e.g., for shaping the highlights and shadows. For example, the default point curve of Medium Contrast is part of this tone shaping. If you adjust exposure, then this affects how the values get mapped to the tone curve.
Thanks for the clarification. Does this mean that if the tone curve is set to linear the exposure slider will correspond to manipulating in-camera exposure or are there other non-linear things (gamma?) going on.
R
-
There are other non-linear things going on with ACR/Lr's Exposure because there are differences between real exposure (in the field) compared to digital exposure. As mentioned earlier, in the field when you reduce exposure you can capture brighter and brighter details. In post-capture digital exposure, you can't. For example, decreasing exposure by 2 stops normally means all your image values get divided by 4. This means what was previously pure white (e.g., 1) is now a gray (e.g., 0.25). And in the field, you can now record new details above that level (e.g., in the range 0.25 to 1). In post-processing, you can't. There's nothing above 0.25. So all you get are "gray" whites, which isn't photographically very useful. ACR/Lr does non-linear stuff to keep whites white (instead of gray).
-
Thanks for the clarification. Does this mean that if the tone curve is set to linear the exposure slider will correspond to manipulating in-camera exposure or are there other non-linear things (gamma?) going on.
I'm not Eric Chan, but if one sets ACR's sliders to linear, the image will still be rendered into a color space with gamma encoding. With Histogrammar, one can take this into account by setting the gamma value appropriately. If one wants scene referred values, the method (http://www.color.org/scene-referred.xalter) suggested by the ICC can be utilized, by converting from ProPhotoRGB to linear_rimm_rgb_ver4. Due to veiling glare, the image will still not be scene referred in the shadows.
Regards,
Bill
-
As of this date one still cannot install Histogrammar v1.2. The expired message still blocks startup.
Is there any way one can overcome this? ...Or?
Thanks,