[argyllcms] Re: collink -g -p white point issue
- From: Graeme Gill <graeme@xxxxxxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Thu, 02 Jun 2016 00:22:00 +1000
Kai-Uwe Behrmann wrote:
The following two commands show a different white point:
collink -i p -g -p abstract.icc input.icc output.icc dl.icc
collink -i p -p abstract.icc input.icc output.icc dl.icc
From the docu I understand, that gamut mapping happens in AbsCol spacewith the -g/-G options. That would give sort of a explanation.
It depends on the intent. A few intents use L*a*b*, most use a CIECAM02
based space in which based on the absolute colorimetric data, but CAM02
makes it close to relative. The Absolute colorimetric type intents
use a common absolute type space.
Note though that abstract profile support is not fully implemented, in that
:- the abstract profile is assumed to be applied after any gamut
mapping, directly in front of the output space conversion, but the
gamut mapping does not take the abstract profiles effect on the
output space into account; the conversion to/from the abstract profile
PCS is done using the output profile viewing conditions if
the link is being computed in CIECAM02 space; if there is a conflict
between the intent the abstract profile is marked with and the intent
of the link, then this is not resolved or warned about (i.e.
what intent is your abstract profile marked with ?)
i.e. I have not attempted to cope with all possible useful abstract
profile usage, nor detect and warn about faulty abstract profile
usage - the facility was only tested to work in the refine capability
in an absolute intent scenario.
How can I workaround the undesired difference between the above results?
Don't include an abstract profile.
Does collink provide a switch to do gamut mapping in RelCol? Would that
Almost all are RelCol or the CAM02 equivalent. Differences in the source
and destination viewing conditions affects the overall conversion though.
-i rl will use standard L*a*b* PCS for linking.
What I'd like is a neutral white point as had expected for perceptual
intent (-i p).
All bets are off when you include an abstract profile in such a scenario
Other related posts: