[Openicc] color-policy vs. color-infrastructure

[Graeme Gill]

Kai-Uwe Behrmann wrote:
> regarding CIE colour spaces. An known issue with them is the lack of gamut 
> information. My concern is: how to preserve the same gamut 

Yes you're right, you can't do colorspace gamut mapping for
CIE colorspaces, without extra information to define the colorspace
gamut.

> What things I could do on application level (CinePaint) to overcome 
> this difficulty with CIE*Lab imagery?
> 
> Would it
> 1. be standard and
> 2. help argyll/lcms
> to add the gamut information found in the device profile to the abstract 
> profile? I am aware of many colour manipulations making the gamut 
> information useless. 
> 
> Other options?

This is certainly one option. Another option (at some slight additional
cost to performance) would be to revert to image gamut mapping (ie. source
gamut is the gamut of the image, rather than the colorspace the image is
defined in.) This would undoubtedly surprise many people used to the existing
way of doing things, and has some subtle issues that should really
be dealt with in some way.

I can imagine additional hybrid modes too, where the source gamut is given by
a collection of images, for instance.

> CinePaint has the color wheels tool, storing its colour manipulation in an 
> abstract profile. This profile type works for 8 and 16 bit depths. I am 
> not shure if an device link could be used across bit depths?

I don't see any reason why not. It would be best of course if
the abstract profile itself was Lab PCS with 16 bit precision.

Graeme Gill.