Is your working space in ACR (see Workflow Options dialog box) set to ProPhoto as well?
Yes, Full Resolution, ProPhoto, 16-bit
I have no explanation for this.
Can you please show screenshots of your ACR dialog showing the color samplers (and their numeric readouts), as well as of your Photoshop window with corresponding samplers & numeric readouts?
I would like to double-check that there isn't a missed setting somewhere.
Hm ... wouldn't it be better to use linear tone curve in the profile, than linearizing default adobe tone curve ?
I have tried both, but shouldn't a tonal value in Photoshop be reasonably similar to the one I chose in ACR?
I'm not sure that I understand what you mean by difference here.
If you mean a small difference between RGB value of the patch #1, which is 242,247,230 by sampling tool and 243, 248,232 as you specified in PS, this is probably caused by different position and/or size of sample between these programs. Don't know about PS, but in ACR, this number isn't a value of single pixel, but average of small area around it. So the small difference is caused by noise in the image
Or you had something else in mind and I missed the point ?
Patch 1 is the only patch that is reasonably close to the ACR output values.
Patch 2 189 175,178,170
Patch 3 144 125,127,123
Patch 4 101 85,86,82
Patch 5 67 53,53,53
Patch 6 37 28,28,28
The variance appears to be much larger than noise. Especially since the avgs in PS are consistently lower than ACR.
I see. Looks like difference between sRGB and PhotoPro values
Anyway, I tried opening similar test raw image, set Photo Pro color space, then opened it in PS, but in my case values shown in ACR and PS are very similar
But wait ... for instance patch #3, you have 127,128,125 shown by ACR tool and 125,127,123 specified on other picture. Where did you get 144 for ACR ?
1 person found this helpful
The 144 number comes from the number on the Point Curve. That's not the correct number to look at, because it is an intermediate value (in the middle of the processing pipeline).
I see now. I will use the numbers at the top from now on. I should also be using a linear profile as well correct?
Just curious, what other processing is being applied after the tone curve?
Thank you very much for you help!
According to dng documentation, applying of tone curve is at the end of pipeline, before converting from Photo Pro to final color space (in your case, there's no such conversion, as you selected Photo Pro as output color space). So I pressume that if you leave all sliders at neutral, that would mean that tone curve is at the end of pipeline in ACR also
Anyway, be aware that you don't see actual tone curve that will be applied. It is a combination of tone curve in the selected profile (which is usually S-shaped), modification of tone curve in ACR Tone curve section and some of sliders in the basic section (brightness, contrast ...)
Couldn't agree with you more, but wouldn't this mean that if I created a custom DNG profile based on a linear tone curve, set all the basic adjustments to 0 or their neutral values, and set no Parametric tone settings that the tonal curve value should match the values once exported to PS CS5?
This is ultimately the workflow I would like to set up.
Well, my knowledge about color science and color management is actually quite limited. Someone with better knowledge about this relatively complex subject would probably give you better or more accurate advice than me
From what I know, typical workflow with digital camera is to transform a capture of the scene to an image that would be displayed on the monitor or printed on the paper. Since dynamic range and color gamut of the scene is usually way larger than of the monitor or paper, image has to be severely tweaked ("rendered") to get pleasant result, while colorimetric accuracy is not very important. This typically involves compressing of the highlights with a tone curve and and some changes of hue and saturation with a big 3D lookup table (+ whitebalancing and changes of color space)
If I understand correctly, what you need is to use your 5D as a kind of scanner, to accurately capture a "scene" that is already on the "output media" with limited dynamic range and color gamut and under controlled lighting. So, these rendering tweaks have to be relatively small, just to compensate for some nonlinearity in the whole process. In other words, tone curve of the whole process should be close to linear and lookup table mustn't introduce huge hue/saturation/brightness shifts
To get linear tone curve of the whole ACR pipeline, you need:
- profile with linear tone curve
- ACR settings: brightness +50, contrast +25, tone curve = medium-contrast
- about blacks slider, default value is said to be +5, while my experiments shown me that it has to be 0, so I'm not sure
- other sliders like saturation, vibrancy, recover, fill light etc should be 0
Couldn`t you just make a custom profile with the DNG profile editor? This should
get you as close as possible to the original colors.
I'm working for as close of match as possible to a fine art original. I have found that creating a custom DNG Profile is only but a small part in this complex process. Even with a proper DNG profile the native tone-curve, brightness and contrast in ACR will cause even a good capture to print badly.
For minimal ACR processing, you should use:
Brightness = 0
Contrast = 0
Blacks = 0
Point Curve = Linear
Sharpening (Amount) = 0
Noise Reduction (Amount) = 0
Everything else at defaults.
This gets you as close to scene-referred as ACR currently permits.
(And be warned this will usually make your images look pretty bad. This is expected behavior.)
Eric, are you absolutely sure about these settings ?
Vit Novak wrote:
Eric, are you absolutely sure about these settings ?
Yeah, I'm pretty sure Eric knows what he's talking about...he works on Camera Raw ya know?
Of course I know, Jeff
I just want to make sure that we were both talking about the same thing, because - default settings for all profiles is brightness 50 contrast 25, and, what's even more strange, my own profiles for Canon cameras with CHDK are giving exact match with jpeg from camera at the very same position of brightness and contrast, and they were calculated with my code, without any usage of DNG profile editor or dng sdk code (where function of brightness and contrast slider are unfortunately not documented), so it was very strange why they don't match jpegs at brightnes 0 contrast 0
The work I've done confirms exactly what Eric said. The emphasis is however on "as close to scene-referred as ACR currently permits".......
I believe that there could be some misunderstanding between me and Eric + there are things in color management that I don't understand correctly
Maybe I ask another question (that posssibliy has different answer): what are the settings in ACR that mean applying no tone curve at all in the whole pipeline, but just conversion from sensor color space to destination color space (which includes gamma encoding to nonlinear output color space, that comes at the end of the workflow, documented in dng sdk) - lIke in some freeware programs, for instance dcraw ?
From a DNG SDK perspective, you can set up the tags in the image such that the pipeline will be in either PCS or scene referred "mode"; what you are talking about is PCS, which sets the processing pipeline not to do the curves, shadow compensation, etc that would be done if the image was scene referred. All raw images are by ACR definition scene referred however.
But that's only useful to know if you're writing code that uses the SDK - it doesn't help the OP's problem.
Hm, it looks like I'll have to study this subject in greater detail ...
However, if I get this correctly, what boswellbk is trying to do is making a dng profile by shooting color checker and then trying to correct tone curve, so that grayscale values produced by ACR, using that profile, match the values specified for that color checker (in Photo Pro color space), at least numbers chosen on tone curve have the same values. Is my guess correct? Is this the right aproach for what he need ?
It is possible to set up either Lightroom or ACR to match the color checker, including the effects of the tone curve - I did so, (actually I did so for Lightroom, C1 and Aperture!) back in 2008 - see the multi-part blog post here. Note that in addition to what I describe in the 2008 articles, you'd also need to use a camera profile without hue twists; those didn't exist back then.
The problem is, as Eric said above, that "accurate" color (aka color without a tone curve) looks terrible!!!!! The original paper on why this is the case, and why the shape of the tone curve needs to be what it is, is "General-Purpose Gamut-Mapping Algorithms: Evaluation of Contrast-Preserving Rescaling Functions for Color Gamut Mapping", Gustav J. Braun and Mark D. Fairchild. Although no doubt Adobe and other have done more recent work.
So I think the original poster is (sorry) on a high road to nowhere in terms of trying to remove the tone curve; it just means the end result will look bad. What can be done is calibrate to the values for the color checker adjusted for a tone curve. That's about as close (in my opinion anyway) as you can get to "scientifically accurate" color repoduction. Of course, you'd also need all the viewing condition stuff done right.
Hm, I'm still not getting this completely, but doesn't matter, I'll read some books about this
Anyway, I'll try to repeat my question in one another form:
Suppose I have a dng profile with a tone curve defined (whether adobe tone curve, or some other tone curve in some Camera profile, or linear tone curve or whatever). What are the positions of ACR sliders that will ensure that THIS curve in profile would be used in the ACR pipeline without any corrections/distorsions/brigtening/darkening/change of contrast/whatever ?
And another question:
If I just want to simulate a scanner with my digital camera, what would be the right approach for this, so that ACR output will match output of a scanner (that has been profiled etc)?
Hi Vit, yes, the settings I described are correct, with the following exceptions:
Hi Vit, I understand your question about the profile's tone curve.
- Contrast = 25
- Point Curve = Medium
Thank you very much for this excelent and clear explanation, Eric. Yes, it is exacty what I have found out so far
I think I understand now why you mentioned brightnes=0, contrast=0, blacks=0, linear tone curve settings to get as close as possible to scene-referred.It is true for a profile with default ACR tone curve (or a profile without a tone curve), as these settings seem to have counter-effect of default ACR tone curve and resulting tone curve of the whole pipeline is linear. However, if I have a profile with custom tone curve, the same settings won't result with linear tone curve of the whole pipeline. That was the reason of misunderstanding I think
Can you confirm this ?
I've been reading this with interest, and it forced be into the color literature where I spent a good part of my morning, getting definitions straight etc. As I look over the comments, it occurred to me that I also had a fine art shoot recently, only one image, but I solved it by using two gray patches about one stop apart, shot under diffuse daylight and set the patches to neutral in ACR. The result was excellent, my client surprised and pleased.
Were I doing a studio shoot in which many different products samples subjects that had to output as high end representations, I would be using the same method you folks do, but with respect to images of objects that do not come close to pushing gamut and tonal ranges (usually!) why this concern with producing a custom profile? I automatically run ProPhoto as my color space. The painting I shot showed zero change when converting to sRGB, using the web viewing function in PS.
This isn't a criticism, I would like to know more about it vis a vis fine art reproduction problems and how this implements the solutions.
You are exactly correct, Vit.
So let me make sure I have this right. If I have created a custom DNG profile with a linear tone curve I choose 0 for my brightness and contrast or the default 50 and 25?
default 50 and 25