Color Editor Tool RGB readings anomaly / Windows 10 /

[AlterEgo]

what their camera profile (Base Characteristics/ICC Profile) RGB numbers correspond to on a Gretag Colour Checker

input profile (camera profile) just defines how raw RGB data that C1 operates on will be mapped into coordinates in a proper color space (PCS) which is either cieLAB or cieXYZ... 

consider a simple case when your input profile is just matrix with PCS = cieXYZ (can't be cieLAB in this case by definition)... matrix M (3x3) describes the linear operation /color transform/ {R, G, B} * M = {X, Y, Z} ... so the code knows which specific coordinates of color in cieXYZ color space those {R, G, B} numbers actually are.

that makes it possible to do further in the pipeline various color transforms into your output space (or you can "embed" your camera profile and postpone the deed) and/or your proof space.

so when you add a new camera you need to add just a new input profile, w/o redefining how its Raw RGB numbers will be mapped into all variety of other color spaces in the universe, that's what PCS and definition of a color transform to PCS is for.

[AlterEgo]

Really, in my experience, it's more important to have a good profile than to match 18 color and 6 greyscale patches out of the hundreds / thousands of actual/different colors in the subject.

and what is the "good" profile ? excluding your "eyes" please.

[Dinarius]

Thanks for the replies.

I have the ICC Profile set to my camera (the default) and View/Proof Profile set to ProPhoto (usually) and Adobe 1998 occasionally.

As I wrote above, a straight (out of the camera) reading from a patch in a Gretag CC will give a Color Tool reading vastly different from the RGB (Proof Profile) at the top of the screen.

However, for what it's worth, if I make a few basic corrections (I'm talking about controlled studio lighting) such as White Balance and Levels, and then take another reading, the RGB values (top of screen and in Colour Tool) are much closer to one another. Also, they are very, very close many of the Gretag patch numbers - as long as I've set the gamma for the image too. (I use a basICColor grey card for this.)

So, happy days, on balance.

Thanks for all the feedback.

D.

[AlterEgo]

But, I know exactly what the RGB numbers are for PP, Lab, 1998, sRGB and Apple. They're readily available.

except that when you have a specific target in your shot then you really don't "EXACTLY" (only approximately) because of variations in manufacturing process (as a minimum) and then whatever numbers for example the manufacturer published... do you really know how they were obtained ? I bet that when X-Rite published new Lab values for November 2014+ CC24 or CCSG targets what happened was that some tech grabbed some i1Pro2 laying around and one of the actual charts (not even several from various batches) and measured (good thing if several times and averaged and not even with different spectometers)... so yes, within several dE2K (<= ~3) those numbers are what they shall be... exactly ? dream on

[AlterEgo]

Thanks for the replies.

I have the ICC Profile set to my camera (the default)

if that is a profile supplied by P1 with C1 and you are not using C1 CH edition camera profiles/curves (as in "linear scientific", w/o instructions for C1 to reconstruct clipped highlights, etc - which makes a difference for areas near sensor saturation) then you can't really hope to get a match (a match let us define as in within 1.x dE2K across all patches vs your measurements of your target... you can't seriously claim a match if your average is > 2.x and max > 3.x )... that profile was not created for that purpose.

[Dinarius]

except that when you have a specific target in your shot then you really don't "EXACTLY" (only approximately) because of variations in manufacturing process (as a minimum) and then whatever numbers for example the manufacturer published... do you really know how they were obtained ? I bet that when X-Rite published new Lab values for November 2014+ CC24 or CCSG targets what happened was that some tech grabbed some i1Pro2 laying around and one of the actual charts (not even several from various batches) and measured (good thing if several times and averaged and not even with different spectometers)... so yes, within several dE2K (<= ~3) those numbers are what they shall be... exactly ? dream on

I can't argue with that. But, there will always be a margin of error. And then, up the line, there is the colour management at the offset printing stage which, as someone who has shot hundreds of catalogues, I can tell you is c**p! 

I can only make it as good as I can. It's then up to someone else to keep it looking that way.

Thanks again.

[Dinarius]

if that is a profile supplied by P1 with C1 and you are not using C1 CH edition camera profiles/curves (as in "linear scientific", w/o instructions for C1 to reconstruct clipped highlights, etc - which makes a difference for areas near sensor saturation) then you can't really home to get a match... that profile was not created for that purpose.

But, I'm trying to match that. I'm trying to match the Proof Profile via what I reading at the top of the screen. I have to say that, on my calibrated monitor, Gretag CCs look pretty darned on the money.

D.

[Bart_van_der_Wolf]

and what is the "good" profile ? excluding your "eyes" please.

Depends on the actual subject colors and illumination conditions. Thus custom made. The CC24 colors are a step in the good direction, especially if one is to make reproductions of, ... CC24s.

Cheers,
Bart

[AlterEgo]

Depends on the actual subject colors and illumination conditions. Thus custom made. The CC24 colors are a step in the good direction, especially if one is to make reproductions of, ... CC24s.

Cheers,
Bart

still how do you actually check that ? as a minimum good process of the profile building shall ensure that you can hit (using the final or intermediate, in case when you then want to fine tune it away from precision towards subjective colors/contrast/etc, profile in raw converter, comparing the actual output from the shot of that actual target with the target measurements using something like babelcolor patchtool) that specific target itself with a certain precision (demonstrated in numbers) - no challenging colors, no metameric issues... but if you can't then it is a matter of subjective opinion only to call it "good"

[AlterEgo]

I have to say that, on my calibrated monitor, Gretag CCs look pretty darned on the money.

and I have to say that you need to use something like patchtool that can eliminate a lot of subjective things, example... I still can't hit this match, which I did with my A7, for my A7R2... I am dE2K = ~1 worse across several patches and stuck there for a few weeks... and _NO_ =  looking @ 2 targets (one from converter, one from spectral data) separately and near each other on my calibrated monitor does not show the difference that looking @ this + actual statistic for dE numbers can illustrate... I do not consider visual match between 2 targets w/o actually intersecting each patch a good match at all...

[The View]

I noticed a really strange thing with color readings.

A model's forehead, brightly lit, only read 35 brightness.

This happened several times. Once it got resolved by double-clicking, and the reading jumped to 230.

Other times I think moving the exposure slider, and back, got the tool unstuck.

It's not the whole image the tool gives wrong read-outs, but weird enough brighter areas of the image, around 210 to 235 - and not all of them.

[Bart_van_der_Wolf]

I noticed a really strange thing with color readings.

A model's forehead, brightly lit, only read 35 brightness.

This happened several times. Once it got resolved by double-clicking, and the reading jumped to 230.

Other times I think moving the exposure slider, and back, got the tool unstuck.

It's not the whole image the tool gives wrong read-outs, but weird enough brighter areas of the image, around 210 to 235 - and not all of them.

Hi,

Have you tried switching hardware acceleration on/oFF to see if that makes a difference in behavior? Also make sure your video drivers are up to date.

Cheers,
Bart

[Dinarius]

Resurrecting this thread.......

Which rendering intent (Edit/Preferences/Colour/Rendering Intent) are people using?

Do you always use the same one?

Thanks.

D.

[Dinarius]

Ps.

Some might be interested in reading how another RAW editing software can be used to address the issue of colour repro, when used in conjunction with a third party plug-in.

D.

[Bart_van_der_Wolf]

Resurrecting this thread.......

Which rendering intent (Edit/Preferences/Colour/Rendering Intent) are people using?

Do you always use the same one?

Hi,

I mostly use Perceptual (like how the output profiles were made), but it depends on the subject and specific image. I might occasionally also try Relative Colorimetric should banding issues in certain gradients appear.

Cheers,
Bart

[Dinarius]

Update:

After an extended ticket correspondence with Support, it is now accepted that, at least with my setup (Windows 10 Pro x64 / C1 Pro), there is a bug in the Colour Editor.

This is after testing it with files from three different cameras. Canon 1Ds Mklll, 5D Mk IV and Sony RX100; all in RAW mode.

With a file in the browser that has NO edits applied to it, if I click on the image with the CE eye-dropper tool, the RGB readings in the CE and top of the screen are identical. Also, the colour swatch in the CE is the same as the colour in the image.

If I make a change to, say, Exposure and then take a sample, the RGB readings in both readouts remain identical.

However, if I make a change to Levels and take a sample, the RGB in the CE is about 20 points higher than the top of the screen readout.

If I make a change to a LUMA curve and take a sample, the RGB in the CE is about 20 points lower than the top of the screen readout.

This is the same pattern for all three cameras.

Editing changes introduce discrepancies in RGB readings, which I'm assuming shouldn't be the case. Support seem to agree with me.

They are not sure when this bug will be fixed.

D.