https://bugs.kde.org/show_bug.cgi?id=486725
Bug ID: 486725
Summary: sRGB content is over-saturated in HDR
Classification: Plasma
Product: kwin
Version: 6.0.4
Platform: Arch Linux
OS: Linux
Status: REPORTED
Severity: normal
Priority: NOR
Component: wayland-generic
Assignee: kwin-bugs-n...@kde.org
Reporter: thefeeltr...@thefeeltrain.com
Target Milestone: ---
Created attachment 169268
--> https://bugs.kde.org/attachment.cgi?id=169268&action=edit
sRGB
SUMMARY
Basically the exact opposite of this bug:
https://bugs.kde.org/show_bug.cgi?id=482780
Instead of colors being washed out in HDR, the colors are over-saturated.
STEPS TO REPRODUCE
1. Enable HDR
OBSERVED RESULT
Colors have too much saturation.
EXPECTED RESULT
sRGB content should look identical in HDR and SDR mode when SDR Intensity is
set to 0%.
SOFTWARE/OS VERSIONS
Linux: Arch Linux
KDE Plasma Version: 6.0.4
KDE Frameworks Version: 6.1.0
Qt Version: 6.7.0
Monitor: InnoCN 27M2V
Driver: Mesa 24.0.6-arch1.2
ADDITIONAL INFORMATION
I have included 3 attachments to help demonstrate.
The attachment labeled "sRGB" is what the image is supposed to look like. This
is with the monitor at default settings with an ICC profile enabled. In a
perfect world it would look like this in all circumstances.
The attachment labeled "SDR Stretched" is what the image looks like without an
ICC profile. This is an sRGB image being stretched to the full gamut of my
monitor (which is around 167% of sRGB).
The attachment labeled "HDR" is the same image with HDR enabled in Plasma
settings with SDR Intensity set to 0%. Since an SDR Intensity of 0% is supposed
to represent Rec.709 it should look close to the sRGB image, but it does not.
It is clearly wider gamut than it should be.
If Plasma were merely stretching sRGB content up to Rec.2020 the "HDR"
attachment would look the same as the "SDR Stretched" attachment, but it does
not. So it is clear that Plasma is attempting to clamp the image down to
Rec.709, it's just not doing a very good job. It lands somewhere in the middle
of where it should be and the full Rec.2020 space.
This is further confirmed with an X-Rite i1 Pro colorimeter and DisplayCAL, as
DisplayCAL applies the ICC profile on its own (which you can see for yourself
by enabling an ICC profile in Plasma, it will get applied by both Plasma and
DisplayCAL simultaneously and the test results will be bad). Since DisplayCAL
is the one doing the color conversions itself it is possible to test ICC
profile color accuracy while in HDR mode. Both tests are run with completely
default monitor settings as to reduce the number of variables.
If you look at "sRGB colors in SDR" image you can see they are basically
perfect, while "sRGB colors in HDR" has them all over the place. While I would
not expect them to be 100% identical they should at least be close.
I do recognize this could simply be my monitor's default settings/handling for
HDR and SDR being wildly different, but I have no way to test this as this
monitor does not allow any adjustments to be made in HDR mode. Which is
partially why I am posting this. I am hoping that someone else with a
colorimeter could also run some tests and see if they have similar results.
If this cannot be replicated, the monitor is probably at fault and hopefully
the issue can be solved once ICC profiles are available in HDR mode. But if it
can be replicated it means it is an actual bug with how sRGB content is being
handled in HDR by Plasma/kwin. Being able to adjust "SDR Intensity" below 0%
could also be useful in this scenario.
--
You are receiving this mail because:
You are watching all bug changes.
