Vectorscope, CIELUV : Missing point in hue ring

[KarlMagnusLarsson]

Introduction

If if cycle the different vectorscopes, the hue rings have six points, except CIELUV, which shows 5 points and a cusp where the 6:th point ought to be. Is this correct?

---

Did you buy darktable from an application store ?

No, I build from source: Darktable git master release-4.3.0-450-gf5a79a849, Linux Debian 11.6.

Describe the bug/issue

If if cycle the different vectorscopes, the hue rings have six points, except CIELUV, which shows 5 points and a cusp where the 6:th point ought to be.

I have had this hue ring issue for quite a while using darktable master source. The picture attached shows a vectorscope CIELUV example with logarithmic scale. The cusp is on the blue side of the hue ring.

To Reproduce

I can reproduce this error for all files (raw, tiff, jpg) for any capture. The make of the camera does not matter (Canon CR2, CR3, Sony ARW etc). All files are affected.

1. Start darktable application and select a raw file in lightroom and go to darkroom.
2. Right click on Toggle gamut checking and set histogram profile = linear ProPhoto RGB (or click toggle soft proofing and set histogram profile from there.)
3. Select Vectorscope CIELUV.
4. See error: CIELUV hue ring, shows 5 points and a cusp where the 6:th point ought to be. The screenshot above in the description shows a typical error.

Expected behavior
CIELUV hue ring, with 6 color points, without a cusp, also for histogram profile = linear ProPhoto RGB. Another case which ought to work is histogram profile = work profile and module settinginput color profile | working profile = linear ProPhoto RGB.

Which commit introduced the error
I use Darktable git master release-4.3.0-450-gf5a79a849, Linux Debian 11.6. However I have had this error for quite a while. The quoted commit is not the commit which introduces the error.

I do not yet have a git bisect commit.
EDIT: I can confirm issue exists in git master tag release-3.8.1.

Platform

• darktable version : git master release-4.3.0-450-gf5a79a849
• OS Linux
• Linux - Distro : Debian 11.6
• Memory : 16 GB
• Graphics card : Nvidia Quadro RTX 4000 8 GB
• Graphics driver : Nvidia 470.161.03-1 amd64 Proprietary Driver
• OpenCL installed : OpenCL 3.0 CUDA 11.4.309
• OpenCL activated : Yes
• Xorg : Yes X11
• Desktop : Gnome 3.38.5
• GTK+ : 3.24.24
• gcc : gcc (Debian 10.2.1-6) 10.2.1 20210110
• cflags :
• CMAKE_BUILD_TYPE : Release

Additional context

• Can you reproduce with another darktable version(s)? The error has been around in git master for quite a while in git master source. I build from git master source. I have not tried precompiled packages.
• Can you reproduce with a RAW or Jpeg or both? Any RAW-file-format/Jpeg/both.
• Are the steps above reproducible with a fresh edit (i.e. after discarding history)? yes.
• If the issue is with the output image, attach an XMP file if (you'll have to change the extension to .txt)
• Is the issue still present using an empty/new config-dir (e.g. start darktable with --configdir "/tmp")? no, it is not present since the default setting works for CIELUV vectorscope hue ring, however setting histogram profile = linear ProPhoto RGB or histogram profile = work profile and module settinginput color profile | working profile = linear ProPhoto RGB will introduce the error again.
• Do you use lua scripts?
  • What lua scripts start automatically?
  • What lua scripts were running when the bug occurred?

[ralfbrown]

I bet the root cause is that Prophoto contains colors which can't be represented in CIELUV - in particular, the blue primary lies outside the visual nexus (the horseshoe on chromaticity diagrams). Rec2020 lies entirely within, so there aren't any colors representable in Rec2020 but not Luv.

[KarlMagnusLarsson]

The ProPhoto RGB blue primary is indeed an imaginary color, but also the green primary is imaginary and the CIELUV hue ring renders the green points on ring perimeter nicely. However perhaps the blue primary is farther away from visible set, compared to the green primary, using some Color difference metric.

https://en.wikipedia.org/wiki/ProPhoto_RGB_color_space

[KarlMagnusLarsson]

I propose to close. The hue ring behavior is a function of CIELUV and ProPhoto RGB definitions. This is also already commented in existing source code.