- This topic has 63 replies, 5 voices, and was last updated 9 months, 3 weeks ago by
.
-
Posts
-
Ok it looks like the XYZLUT profiles i’ve been creating also come with 3 curves and a matrix, and KDE must be just using those.
Do XYZLUT profiles also come with VGCT curves?
I depends on if you set tone responce:as measured (No VCGT) or a different option(VCGT).
I’ve been using -aX with colprof
https://www.argyllcms.com/doc/colprof.html#aReading through that, I just now learned about the existence of -aY which I used to verify that KDE is indeed ignoring the cLUTs.
So, to confirm the behavior of KDE/Wayland:
- Uses the matrix profile. Added a simple profile that inverts red and blue for testing.
- Uses the TRC curves (corrects your gamma and green cast).
- Uses VCGT (from the profile I attached) to correct gamma and the green cast. But you cannot use DisplayCAL to create a VCGT under Wayland.
- Does not use 3DLUT.
Yes, your profile works and your observations are correct except for the following:
DisplayCAL *can* make any ICC profiles just fine, it just crashes before it ever gets to that step due to a bug.
If colprof is run with the “-d” option set, it causes KDE to interpret the profile strangely and causes my original issue. I dont know what option in DisplayCAL caused the “-d” option to be set, and to test it I would have to spend some arbitrary amount of time between 1 and 6 hours for each calibration to get the DisplayCAL crash to not trigger. I assume this fixes the crash because it doesn’t show you that screen that asks you to install the color profile (which i think is the cause of the crash) if you spend a very long time calibrating/profiling. I submitted a bug report about it yesterday and I’m waiting for it to get noticed.I initially left this option as the default since it didn’t seem entirely relevant, but it turns out that if you change the option, the help message changes too, which is an odd design choice.
Efficiency mode (matrix + shaper ONLY):
colprof -d doesnt workAccuracy mode (allows XYZ cLUTs):
colprof -d AND gamut mapping doesn’t work(I assume gamut mapping requires the use of cLUTs)
Ok I just set and unset the ICC profile and now the gamut mapping one is working.
It’s really inconsistent…
Do XYZLUT profiles also come with VGCT curves?
Display profiles have two parts:
1-grey calibration = VCGT, which could be linear ramp fo HW cal profiles or default vendor profiles or no calibration profiles.
2-colospace definition, which could be
2.a simple: matrix with RGB primaties location + “gamma/TRC”
2.b mesh like, like in XYZLUT profiles or Xrite “table” profiles. DisplayCAL XYZLUT profiles also stored a matrix simplified gamut description for compatibility, although some apps may reject to color manage at all with XYZLUT.XYZLUT profile IS NOT a LUT3D. Is a mesh describing display behavior. A LUT3D is a mesh, precalculated transformation between 2 profiles. An ICC profile cannot store any LUT3D because it does not know source colorspace (content colorspace: an sRGB JPG, and AdobeRGB TIFF, …)
ICC color management modules do not use LUT3D but likely compute this transformation of the fly depending on content colorspace, while video editors ro players… whcih are likely to be using same content colorspace all teh time (Rec709) usually offer the option to use this prcalculated transfromation (a LUT3D) whcih can run in GPU using all its acceleration .The colorant and TRC tags desctibe the device behavior and are used to compute corrections on the fly.
The BTOA tag however stores an actual 3DLUT, along side a matrix and input/output curves for the purpose of converting from linear XYZ to nonlinear RGB (for monitor profiles).
The displaycal profiles attached here set the matrix to unity and use the 3DLUT to map XYZ to RGB.
Other apps follow a different approach like using the matrix to convert XYZ to RGB and the 3DLUT to finetune the RGB.
Further indication that these 3DLUTs are used for interpolation and not characterization is the fact that the inverse transform is stored in a separate tag, not expected to be computed on the fly.
I’d bet he was talking about “RGB to RGB” LUT3D, hence my clarification. So there is no LUT3D (RGB2RGB) in his ICC profile because profile does not know the content colorspace.
