Hi,
Although I am not the original poster, I did recognize this 'issue' from the
past.
I just tested your beta code, Graeme.
It does a wonderful job. Now, the results are as expected, as far as I am
concerned.
Thank you!
I am curious about Sebastian's results...
Greetings,
Auke
Graeme Gill wrote:
Sebastian Parborg wrote:
Hi,
Here is all the generated files besides the .tiff file of the test
pattern.
https://drive.google.com/file/d/1ZG2Q6Cxvj6F5lq9tiubq3AwjPIzrSXX9/view?usp=sharing
<https://drive.google.com/file/d/1ZG2Q6Cxvj6F5lq9tiubq3AwjPIzrSXX9/view?usp=sharing>
thanks for the details.
I've had a look at it, and think this is an artifact of the imperfections
in the CIECAM02 based appearance space that ArgyllCMS uses for computing gamut
mappings and clippings. While CIECAM02 is far better than L*a*b* in this regard,
it still seems to have issues with the hue linearity of saturated blues.
This is the reason it tends to show up with mapping from additive colorspace
highly saturated blues to subtractive (i.e. print) colorspaces with particularly
small or rounded gamuts near blue. (A gamut that has a prominent blue cusp will
tend
to draw mapped/clipped blue colors towards it, possibly counteracting the
effects
of colorspace hue non-linearity.)
The IPT and other successor perceptual spaces seem to be better in this regard.
Ideally I'd like to replace CICEAM02 in ArgyllCMS with something mathematically
simpler and more up to date, perhaps based on the CAM described in the paper
"A Neurophysiology-Inspired Steady-State Color Appearance Model" by Timo Kunkel
and Erik Reinhard, but that's a larger project than I have time for at the
moment,
so what I've done is to add a hue linearization tweak to the CICCAM02
implementation
inspired by the paper "Color gamut mapping in a hue-linearized CIELAB color
space"
by Gustav Braun and Mark Fairchild. This selectively modifies the Hue of highly
saturated blue colors to better line them up visually with their less saturated
counterparts.
Beta test MSWindows 64 bit colprof is here:
<https://www.argyllcms.com/win64_V2.2.2beta_colprof.zip ;
<https://www.argyllcms.com/win64_V2.2.2beta_colprof.zip> >.
Source code snapshot is in the usual place:
<https://www.argyllcms.com/Argyll_dev_src.zip ;
<https://www.argyllcms.com/Argyll_dev_src.zip> >
Give that a go and see what you think.
(If you need an executable for a different platform, let me know.)
Cheers,
Graeme Gill.
