Color Difference after profiling

So I noticed that even if I have both XNView MP and darktable set to use my Asus monitor profile for displaying, XNView the image looks a lot darker:

These are my settings:


Does anyone know what could be the problem? The photo looks a lot better (almost the same) when opened in a browser. In any other image viewer it looks just like in XNView…

Re-opening the exported photo in darktable makes it look just like darkroom RAW.

If I set darktable view intent to “relative colorimetric”, then both the histogram and colors match a bit closer what I see in XNView.

Last thing to note… the Asus profile is installed as system default on Windows 11, if it helps…

Xnview has settings that make a difference…or can ie the relative and perceptual and there is a bp compensation setting…I don’t think generally in DT they make a difference, but with your custom profile you may have made your display profile with bp compensation… There were a set of test images that you could use to show the rendering intent that was being used…

So make sure xnview and DT are using the same display profile and then check those settings in xnview…if I get a chance I will check which ones I se that I feel looks the same…

I tried them and didn’t see any difference to be honest…

I did, I tried multiple combinations and it never looked the same.
The closest I got was setting DT display to relative colorimetric while keeping the rest perceptual :thinking:

Thanks.

I’ve read that, if a matrix-type profile is set to ‘perceptual’, then that is ignored by a browser or viewer which instead renders ‘relative colorimetric’. I don’t know what dt does in that regard.

Do you mean your calibrated DIsplay cal profile when you say your asus profile??

Here is the test page…

https://displaycal.net/icc-color-management-test/

https://displaycal.net/icc-color-management-test/

If you check the win11 photo app you can see its different than xnview at least how it handles the test image…

The only rendering intents that work in DT as far as I know is if you have a printer profile for softproofing or an icc profile that has a lut when you export…the DT default profile is a matrix profile…the test images indicate what gets picked up when you open them in an app ie what profile is used if your profile has both a matrix and a lut…

I think I went through all this in a thread if you search you might find it…I have to run and make dinner…I can check later

EDIT

I also noticed on xnview you sometimes need to go to a new image and back after you change the settings to see the changes…

I looked closer…if anything you should softproof that the export profile not your display profile…that can if the differences are enough show you what you are likely to see in other software as that is the embedded profile not your display profile…

1 Like

It’s past midnight here, I will check tomorrow (well, technically, later today :upside_down_face:) . I’m guessing you live in the East/Central US?

1 Like

I suspect that profiles are at the heart of the issue.

For example here is a photo taken from a recent Play Raw with one conversion in the GIMP (top left) and three different assignments (not conversions):

The bright ones are RIMM gamma and linear; the bottom [right] one is perceptual V4 with CLUTs.

Perceptual rendering intent requires a profile with a LUT color transform.

I knew that.

All except top left have CLUTs (AtoB0 etc).

Yes, from now on, I will refer to it as the “Asus DisplayCal profile”

Oh and btw., I refer to XNView MP as XNView, for short

That makes the DT even lighter, but I guess that’s supposed to happen.

The test images are handled well in XNView if I use e.g. my laptop DisplayCal profile, but not with my Asus DisplayCal profile.

So the profile itself seems to be the culprit then. Because none of the test images are displayed correctly in XNView with it and it looks the same as color management turned off (confirmed by the test).

Swapping the viewing profile both in DT and XNView for something else (like the sRGB Appearance or even my laptop’s DisplayCal profile) → that gives me a nearly perfect match between the apps.

What now? Can I fix the monitor profile or what do I even do…? I made the profile by closely following the tutorial from @Donatzsky.

A bit of an offtopic, but where can I find some test images with built in profiles?

I’m changing my image viewer from qcms to lcms, for future 16bit support etc, and would like to test for correctness. I’ve used ICC Color Management & Rendering Intent Test (all passing :slight_smile: ) but would like to have more on hand.

Thanks for the link! Duly downloaded …

1 Like

Just a bit of over-arching context: you should treat whatever render you get on your display with a profile calibrated to that display as “gospel”. That’s the whole point of a calibrated display profile, to produce image data in a render specifically tailored to the tone and color capabilities of that display. Get that to work right, and measure any other way in terms of that anchor.

Otherwise, it’s easy to get into a “chasing squirrels” sort of meander.

2 Likes

Well what now :slight_smile: I went to check the guide…I guess one thing you could look at is what you set for BP compensation…I made mine and turned it off. I found I got deeper blacks which I liked and I don’t have OLED screen tech so I like the look and I was always using srgb or 2.2 for gamma but I read a few places where it actually suggested using as measured I think or what ever option uses the native gamma of the monitor…something about its easier for the profile to not have to compensate for gamma too much…not sure when I checked it seemed like the native gamma was approx 2.3 for my monitor so I tried the as measured and made my profile that way…don’t take this as advice. If you are curious I can try to track down a proper explanation for this sort of approach… I don’t think I did bother to compare setting it to native vs 2.2 I just went by the report which was fine and my visual assessment which I was happy with…I don’t know if that color appearance profile is helpful other than it allows for you to choose a rendering intent. I did like the nice contrasted output it produces in some cases for edits but I think using the standard export profile in DT might generate a closer match to what you see in DT… so I guess it would fall on you to decide the route to go there…

1 Like

I don’t remember it clearly, but I’m 99% sure that I left black point compensation unchecked. Is there any way to check what I selected? Does this help?

That’s exactly what I aimed for, it’s just that the profile is not working in XNView. It’s either the software acting up (unlikely) or the profile is somehow wrong… But darktable accepts it, and seems like Vivaldi browser does handle it just fine as well. What’s going on :crazy_face:

*I should probably note that both profiles I made yesterday don’t work in XNView, so they might have something in common… But the profiles I made last november do work… HUH?

If it is of any help, here they are:
This one works:
PA248QV #2 2024-11-24 19-22 2.2 F-S XYZLUT+MTX_sRGB_gamma2.4.icm (946.2 KB)
This one doesn’t:
PA248QV #2 2025-03-28 19-45 125cdm² D6500 2.2 F-S XYZLUT+MTX_sRGB_preset.icm (946.4 KB)

The problem is the older one doesn’t cover sRGB as well and I don’t exactly remember my monitor calibration settings for it, so…

1 Like

Could be XnView MP

I have downloaded the ICC test and opened all six test images in XnView … not one of them passed.

In FastStone Viewer, four passed but not the PNGs

I would vote for “the profile is somehow wrong” …

Exactly my experience - I guess it doesn’t support color profiled PNGs then, thanks for feedback.

I wonder what could be the issue, maybe this? I used the “Auto (none)” option before and I don’t think the settings in the Calibration tab matter… Otherwise not much changed in the way I made the profile as far as DisplayCal itself is concerned
image

The images have some weird behaviour…in both XNview and DT if you load the jpg files with the clut only icc and the one with the matrix fall back…they will fail but it you zoom to 100% and in the case of DT enable HQR…then they don’t …so something in the setup there I think cant’ be trusted with these images…

If you are talking about the ICC test, the images are deliberately designed to fail if there are problems with an app’s color mamagement.

So I am not sure what you mean by “can’t be trusted”, please clarify …