Hardware calibration, how to set Gnome Color/Darktable ?

color_management

(Aurélien Pierre) #1

Hi,

I have a BenQ PG2401 photo monitor. This device has hardware calibration, meaning that (on Windows), whenever I calibrate it (with the BenQ rebranded xRite soft), the LUT should be flashed into the monitor chip (which has to be connected in HDMI + USB), and no color correction should be made on the computer/GPU side.

Now, back on Ubuntu Gnome 16.04 with Darktable, should I let the monitor color profile blank or set it to Abobe RGB ?

I ask this because I ran into a strange problem yesterday. I just moved from the Unity desktop to Gnome. Retouching a low-key picture in DT was fine with AdobeRGB as the monitor profile, but the “system” monitor profile (set on default on the OS side) seems to have a black-point compensation issue :

Similiarly, the exported TIFF file with Adobe RGB profile looks good in Gnome viewer and Photoshop :

So… I’m a bit lost between the inputs and outputs.

Anyway, the final image in sRGB is here : https://photo.aurelienpierre.com/maquillage/


Image viewer for linear gamma images
#2

When you set a LUT inside the monitor you must have specified some well known color profile which the monitor should emulate. If that was AdobeRGB then yes, set AdobeRGB as your display profile in the applications.
You might want to check if the profiling/calibration software on Windows created a system display profile (some .icm file) and use that instead. It might have some fine tuning embedded or at least a linear VCGT so that the risk of double correction is less likely. In general both should be fine, either a standard AdobeRGB profile (which doesn’t have any VCGT by default) or a display profile created on Windows.


(Aurélien Pierre) #3

Using the .icm profile generated by xRite gives the same problem of shadows becoming solid blacks. That’s why I wonder. But I didn’t check against which color space it was produced (although I set everything in AdobeRGB). I will check it when I reboot.

Here it is : BenQ PG2401.icm (13.0 KB)


(Aurélien Pierre) #4

The color space was AdobeRGB indeed, but the profile was ICC V4, which is not supported by Argyll. Redoing an ICC v2 solved it for DT. But now Firefox is messed up with over-saturated colors.


#5

Salut, Pierre,

Firefox -> about:config -> gfx.color
What does it say?

/Claes


(Aurélien Pierre) #6

Hi Claes !

gfx.color_management.mode = 2
gfx.color_management.rendering_intent = 0
gfx.color_management.enablev4 = false

But if I set the system color profile to Adobe RBG color space, everything is alright. I guess using the hardware profile applies twice the same corrections.


#7

No string value in gfx.color_management.display_profile?

/Claes


(Aurélien Pierre) #8

No.


#9

(Aurélien Pierre) #10

After setting the browser as suggested by @Claes and @Elle, the pictures look the same.

But still, with the monitor color profile on the OS side, I get more saturation on Firefox

It’s fine with the /usr/share/color/icc/colord/AdobeRGB1998.icc profile though.


#11

Hmm, encore…
What icc do you use?
Have you made it yourself?

/Claes


(Aurélien Pierre) #12

BenQ PG2401.icm (1.3 MB)

It’s done by xRite/BenQ Palette master.


#14

Merci, Pierre!

I can check later tonight (hoping that it would make me wiser).
In the meantime, have you played with Elle’s
https://ninedegreesbelow.com/photography/color-management-experiments-1.html

/Claes


(Pat David) #15

I’ve moved your posts into the appropriate topic for you. :slight_smile:


(Elle Stone) #16

I never responded this thread because it involves things like “Unity”, “Gnome”, “colord”, and “Ubuntu 16.04”, plus monitor profiles made in Windows using proprietary software and flashed into the monitor chip. However, today I:

  • Read through this thread and noticed the LAB LUT BenQ monitor profile

  • Installed an ArgyllCMS LAB LUT monitor profile as my own system monitor profile and told Firefox to use this LUT profile, and similarly with GIMP

  • Fired up Firefox and looked at some images in my online Gallery - the colors are horrible.

Having done the above, I can see some very clear reasons why @aurelienpierre might have formed a very bad opinion of ICC profile color management.

Firefox hasn’t supported the use of LUT monitor profiles since Firefox V4, which was a long time ago. Back then Firefox used LCMS as the color management engine. LAB and XYZ LUT monitor profiles were supported. Black point compensation was supported. Photographers were happy.

Then a security alert was issued for LCMS and instead of waiting the short time for a fix (like Firefox never has its own security issues), Firefox decided to roll their own color management system, “qms”, and Firefox lost support for LUT monitor profiles and also for black point compensation.

According to this bug report filed 9 years ago, the lack of Firefox support for LUT profiles was “fixed” 6 years ago:

But the “fix” hasn’t yet been committed to Firefox code:

The last comment to the above bug report was three years ago. So it’s anyone’s guess if Firefox will ever support LAB LUT monitor profiles. As I recall, for XYZ LUT monitor profiles probably Firefox only used the embedded matrix if any instead of the actual LUTs, and if not uses sRGB. Leastways this is what Google Chrome does.

Which brings the topic around the problem of ICC profile implementation. Every OS and application seems to do their own thing, leaving users to deal with the resulting inconsistencies.

I’ve dealt with this issue of inconsistent implementations over the years, bit by bit, and never really stopped to think about all the things I’ve done to ensure that ICC profile “stuff” on my own computer produces consistent results across applications.

Anyway, some time tomorrow I’m planning to make a pixls post entitled something like “Comparing OCIO and ICC color management and implementation problems”, or “25 things I don’t like about ICC profile color management, that maybe OCIO does better”, or etc, unless perhaps @aurelienpierre starts a similar thread first :slight_smile: . It would be very nice for ICC profile users to have a working understanding of OCIO and vice versa, and I’m guessing there is less room in OCIO for variations in implementation.

In the meantime, @aurelienpierre - did you ever manage to get darktable and your other softwares to display consistent results? If not, I think @gwgill might have insight into navigating some of the complications resulting from using colord and GNOME, but then again your version of Ubuntu might still use Mir, that might complicate things.

That huge LAB LUT BenQ profile provides all six AtoB/BtoA tags. My own ArgyllCMS LAB LUT profile only has a single AtoB0 tag, for relative colorimetric intent, which somewhat reduces the possibilities for “different results on different softwares and operating systems”.

Apart from affecting the black point, and depending on the conversion intent and whether black point compensation is used (implementation of both of these vary from one ICC profile application and CMM to the next) that BenQ LAB LUT profile does seem very close to AdobeRGB1998. But I say this based only on looking at a couple of images and converting to the LUT profile and then assigning AdobeRGB1998. I haven’t taken the time to use ArgyllCMS command line utilities to plot the respective color gamuts and TRCs of AdobeRGB1998 and the BenQ LAB LUT profile.


(Graeme W. Gill) #17

There’s really no reason for any open source or free software application using ICC color management to use anything other than lcms.

Commercial applications will tend to use the operating system CMM (OS X, MSWindows) or use their own thing if they have the prowess (i.e. Adobe.)

FWIW lcms and ArgyllCMS are in pretty good alignment with Adobes CMM behavior. The operating system CMM’s seem to get less attention than they should in regard to conformance to expected norms though.

Applications mis-applying color management and/or not providing good transparency about how they are applying it seems to be par for the course. Many programmers, even those heavily involved in computer graphics seem to be largely ignorant of color science and how to apply color management.