Edge Chromium seems to be color managed

[rasworth]

Loaded up Edge Chromium on my W10 system, it's a giant step up from Edge.  I used it to view several images with various profiles, displayed them visually identical to Photoshop.

I have a wide gamut monitor, profiled to a gamut similar to Adobe 1998.  Edge Chromium has a picture of the day function, when I first invoke it the image displays over-saturated similar to the old Edge, but after several seconds the image "snaps" to decreased saturation, again seems visually correct.  Somewhat strange behavior, but certainly much improved.

Richard Southworth

[kers]

Loaded up Edge Chromium on my W10 system, it's a giant step up from Edge.  I used it to view several images with various profiles, displayed them visually identical to Photoshop.

I have a wide gamut monitor, profiled to a gamut similar to Adobe 1998.  Edge Chromium has a picture of the day function, when I first invoke it the image displays over-saturated similar to the old Edge, but after several seconds the image "snaps" to decreased saturation, again seems visually correct.  Somewhat strange behavior, but certainly much improved.

Richard Southworth

So then they are almost there after..... Years

[GWGill]

Loaded up Edge Chromium on my W10 system, it's a giant step up from Edge.  I used it to view several images with various profiles, displayed them visually identical to Photoshop.

Are you sure ? Reports indicate that Chromium color management code is currently broken, and only implements the matrix part of a matrix profile on X11 and MSWindows and substitutes its own per channel curves. And it doesn't do cLUT profiles at all.

See this and this.

[rasworth]

Am I sure beyond a reasonable doubt? No, but with the few images I tried I found no problem.  Perhaps you could suggest a more rigorous experiment.

Richard Southworth

[rasworth]

I loaded an image with ProPhoto embedded, into Firefox and Edge Chromium side by side, and took a screen shot (attached).  Again, without any measurements, the image appears identical.

Richard Southworth

[rasworth]

A similar screenshot comparison, Edge Chromium left and Photoshop right, again using an image with ProPhoto embedded.

I prepare my screenshots by pasting into a new Photoshop image, flattening layers, assigning the monitor profile, and then converting to sRGB, for reasonable display on most monitors.  The intent is to create a visual comparison, not present accurate colors.

Richard Southworth

[GWGill]

A similar screenshot comparison, Edge Chromium left and Photoshop right, again using an image with ProPhoto embedded.

You need to be testing different display profiles though, not source profiles - i.e. cLUT profiles and/or a  matrix profile with a non-sRGB like set of per channel curves.

[fdisilvestro]

I have the same perception about chromium. Attached is a quick comparison between Chromiun and Firefox, using a wide gamut monitor.

[rasworth]

GWGill,

I don't understand, I have a system wide display profile loaded for my wide gamut monitor.  I have observed that non-color managed applications that don't utilize the monitor profile generally display overly saturated images.  Since the test images display correctly, i.e. identically to a fully color managed application such as Photoshop, then I'm assuming Edge Chromium is properly using my monitor profile (matrix based).  In what way should I be testing different display profiles?

Richard Southworth

[GWGill]

I don't understand, I have a system wide display profile loaded for my wide gamut monitor.  I have observed that non-color managed applications that don't utilize the monitor profile generally display overly saturated images.  Since the test images display correctly, i.e. identically to a fully color managed application such as Photoshop, then I'm assuming Edge Chromium is properly using my monitor profile (matrix based).  In what way should I be testing different display profiles?

As I explained above, the previous reports about Chromium were that it was rejecting cLUT profiles, and only using the matrix from a matrix profile (on X11 and MSWindows versions. On OS X versions apparently the full matrix profile was still being implemented). Now it's possible that new versions of chromium have fixed the problem. But saying that images on a Wide gamut monitor "look OK" doesn't really prove it.

The most obvious thing about a wide gamut monitor when it is not color managed is over saturated colors. Just implementing the matrix part of a matrix profile largely corrects this. But if your display has an sRGB like response curve, and chromium is substituting its own curves for the ones that are actually in the matrix profile, you may be unable to notice this.  It's only if your display has a luminance response curve that is visually different from sRGB that you will notice the discrepancy as the chromium image looking too dark or too light.

A possible way of testing would be to manually alter you display to have a non sRGB like luminance curve. You'd need to figure out which of your display controls (i.e. brightness, contrast, or better yet if it has a gamma control) changes the luminance curve without changing the white brightness much. Then alter this so that the display looks obviously wrong (too dark or too light),  and the profile it while skipping calibration. (You would need to make absolutely sure that your profiler doesn't change the display controls or create calibration curves). Double check that the matrix profile curves (i.e. rTRC, gTRC & bTRC tags) are different by examining your old and new profiles, and the do you comparison of chromium vs. a known color managed application.

And of course unless you try a cLUT profile you won't know if they work or not.

[rasworth]

GWGill,

I never claimed "proof", only that it visually appeared to be color managing correctly.  And you are correct, when I calibratred/profiled my monitor I chose gamma = 2.2, which would put it very close to sRGB.

I'll leave it to someone more ambitious than me to perform your experiment.  Thank you for your detailed explanation.

Richard Southworth

[rasworth]

Changed my mind, did what GWGill suggested - created a display profile with gamma = 1 instead of 2.2.  The first observation is this makes for a really ugly display for non-color managed images.  Photoshop and Firefox both displayed images (with embedded profiles) correctly, but Edge Chromium flunked.  As GWGill predicted, it did not pick up the tonal curves from the monitor profile, and of course the resulting image looked like a non color managed image, i.e. very flat with gamma = 1.

So Edge Chromium is not completely there, wrt color management.  Practically speaking it may not make much difference, since most monitor profiles are created with gamma = 2.2, which is apparently close to the tonal curve Edge Chromium is forcing.

Why do they come this close and then not take the last step?  Does this somehow make the product more universal, or is it just programmer laziness?  Somebody please explain.

Richard Southworth

[digitaldog]

Practically speaking it may not make much difference, since most monitor profiles are created with gamma = 2.2

I don't know that's actually the case and there are many cases where it's not ideal and lastly, IF Edge Chromium can't deal with anything but profiles built to (gamma 2.2, not TRC 2.2), it's broken, not "partially color managed".

[rasworth]

Yes Andrew, you're absolutely correct.  At least give me credit for following thru with the experiment.

Richard Southworth

Added by edit - any idea why they half-assed it?

[digitaldog]

Yes Andrew, you're absolutely correct.  At least give me credit for following thru with the experiment.

Richard Southworth

Added by edit - any idea why they half-assed it?

Yes credit for the work thus far.
Yes. Half assed because it's not properly and fully color managed.

[rasworth]

You gave me a tautology ("the saying of the same thing twice in different words, generally considered to be a fault of style").  I'll restate, can you think of any advantage to be gained doing it with a hijacked tone curve?  It seems to be too deliberate to just be an error.

And yes, I realize you know the definition of tautology, just yanking your chain.

Richard Southworth

[digitaldog]

You gave me a tautology ("the saying of the same thing twice in different words, generally considered to be a fault of style"). I'll restate, can you think of any advantage to be gained doing it with a hijacked tone curve? 

I don't see any advantage in hijacking anything. 

[FabienP]

You gave me a tautology ("the saying of the same thing twice in different words, generally considered to be a fault of style").  I'll restate, can you think of any advantage to be gained doing it with a hijacked tone curve?  It seems to be too deliberate to just be an error.

And yes, I realize you know the definition of tautology, just yanking your chain.

Richard Southworth

Since the development of the Chromium browser (which is the base of Chrome and of the new MS browser also known as Chredge or Edgium) is open sourced, someone cleverer than me should be able to retrieve this bit of information. I have tried and did not manage to find something useful.

Speculating a bit, I could see two reasons to do things they way they were done in the current implementation:

• The developers are not really aware of what should be done to fully support colorimetry in their browser and mostly succeeded despite this imperfect knowledge.
• The developers found a way to achieve 90% of the goals of the feature request with 10% of the required effort by hardcoding some things and making some assumptions (that aren't always met) that gamma is always set at 2.2. This solution is good enough and will make it possible to work on some other stories in the same code sprint.

In my opinion the latter case would be the more pleasant explanation.

Cheers,

Fabien

[kers]

As soon as MS sees a way to make money by implementing colourmanagement in the browser the problem will be solved.

[GWGill]

Does this somehow make the product more universal, or is it just programmer laziness?  Somebody please explain.

Programmer laziness. Someone reported some problems, and rather than investigating or understanding much about color management (or even try and ask someone who might know), they simply turned bits of it off, and "blamed Windows".

The basic logic goes something like this: "We're risking lots of bug reports by not doing something; no-one uses color management, so it's a good tradeoff to cripple it while still adressing wide gamut issues".