Re: [Gimp-developer] Adding better LCH support to GIMP 2.10

[Elle Stone <ellestone ninedegreesbelow com>]

On 02/14/2016 03:56 PM, Sven Claussner wrote:
> Hi,
>
> @Elle: you are speaking here of the JCH color model (or space) and
> mentioned
> on your website [1] that JAB and JCH have outdated LAB and LCH
> (which are often considered the high end image editing color
> spaces/models).
> Searching a while for more information about JCH I found only very few
> information, even not on other color management and FOSS graphics devel
> mailing lists. Only on the PXLab website [2] I see a short description:
>
> JCH:=The CIE Color Appearance Model (1997) with viewing and scene
> conditions
> to be defined separately.
>
> I'm failing to understand all its implications.
> Can you tell us more about JCH and JAB and why you consider it to
> be a good choice, please? What about LAB and LCH then?

I can't tell you very much about JCH/JAB because I'm still trying to 
figure it all out myself.

You might try plowing your way through this PDF:
http://rit-mcsl.org/fairchild/PDFs/AppearanceLec.pdf

This page has links to some equations: 
https://en.wikipedia.org/wiki/CIECAM02

This page briefly talks about color appearance models in more 
down-to-earth terms: 
http://www.rit-mcsl.org/fairchild/WhyIsColor/Questions/4-8.html

An internet search on terms like Color appearance model, Mark Fairchild, 
and CIECAM02 will turn up a lot of material. None of it is easy reading.

LAB answers the question "how far apart do colors have to be before the 
average human observer will say 'those are different colors'". The 
"home" of LAB was for use with quality control for colors in textiles, 
printing, and such. LAB wasn't designed to be used as a color space for 
editing, but it works pretty well for a lot of different editing tasks.

A full understanding of LAB would require understanding the kinds of 
experiments that were done to map out "when is color X different enough 
from color Y to be seen as visually different?" I have no idea what kind 
of experiments were done or how the mathematical model was constructed 
from the experimental results. But the resulting equations to convert 
from XYZ to LAB and then LCH are pretty straightforward.

Color appearance models are designed to answer a very different and much 
more complicated set of questions. They try to answer questions like 
"How to describe colors?", "Why does the appearance of one color change 
when juxtaposed next to another color?", and "Why does a surface look 
the same color even when the light shining on it changes drastically 
(for example from bright daylight to deep shade or to tungsten lighting, 
or even in the shadow side of an object)?".

These are complicated questions, being answered using complicated 
research and complicated resulting models described by complicated sets 
of equations. Also color appearance models are a very active area of 
ongoing research, so what's considered really good today might be 
superceded tomorrow.

My reasons for suggesting that for use in GIMP (1)LAB/LCH is good and 
(2)JAB/JCH is probably better are so simplistic that you'll all just laugh:

1. Bruce MacEvoy's handprint.com website on watercolor pigments switched 
from using LCH to using JCH to give paint pigment colors, and I respect 
Bruce MacEvoy as an authority on giving useable values for paint 
pigments: http://handprint.com/HP/WCL/water.html, 
http://handprint.com/LS/CVS/color.html

2. Mark Fairchild seems to think CIECAM02 is a pretty good color 
apperance model, though again this is a field of active research, and 
much of what Mark Fairchild writes goes right over my head.

3. ArgyllCMS and LCMS already incorporate the equations for JAB/JCH in 
their code, as does RawTherapee, so it has to be not too difficult to 
code up for GIMP. But my (very limited) efforts to make sense of the 
ArgyllCMS/LCM/RT code have so far been unsuccessful.

I find the LCH blend modes, color picker, and Hue-Chroma tool to be 
incredibly useful. The few people who've written to me about using my 
patched GIMP seem very enthusiastic.

JCH apparently is more accurate than LCH for describing colors. I'm not 
sure how much difference the "more accurate" would make in the digital 
darkroom. It would be nice to be able to give JCH a try, but in the 
meantime I can't imagine going back to editing without LCH.

> Does somebody else here know more about it?

No doubt even my short description above needs a lot of corrections!

> Thank you in advance
>
> Sven
>
>
> [1]
> http://ninedegreesbelow.com/photography/high-bit-depth-gimp-tutorial-edit-tonality-color-separately.html
>
>
> [2]
> http://irtel.uni-mannheim.de/pxlab/doc/api/de/pxlab/pxl/ColorSpaceCodes.html#CS_JCh
>
>
> _______________________________________________
> gimp-developer-list mailing list
> List address:    gimp-developer-list gnome org
> List membership:
> https://mail.gnome.org/mailman/listinfo/gimp-developer-list
> List archives:   https://mail.gnome.org/archives/gimp-developer-list