[argyllcms] Re: profile -S -g
- From: Klaus Karcher <lists@xxxxxxxxxxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Thu, 24 Apr 2008 22:16:49 +0200
Klaus Karcher wrote:
profile -S 'ISOcoated_v2_eci.icc' -cpp -dpp solventLFP
I get *heavy* clipping when using the perceptual intent.
Graeme Gill wrote:
That's most surprising. The source gamut is obtained from
the colorimetric A2B table. It should not be subject to any
gamut mapping, and should represent the true gamut of the device.
sorry, maybe I was not clear enough.
cctiff -ir ISOcoated_v2_eci.icc -ip solventLFP.icc
does exactly what it shold.
cctiff -ip ISOcoated_v2_eci.icc -ip solventLFP.icc
is the one that causes clipping.
... but unfortunately I can use neither device links nor "mixed"
rendering intents with the RIP.
How about rebuilding the ISOcoated_v2_eci.icc using Argyll ?
Definitely a good idea ;-) ... but no solution for my problem: the
workflow must honor embedded profiles.
The attempt to build a profile that uses the *expanded*, "virtual"
iC2-Gamut to the solventLFP device gamut fails:
iccgamut -ff -ip -cpp ISOcoated_v2_eci.icc
mv ISOcoated_v2_eci.gam iC2pf.gam
profile -S ProPhoto.icm -g iC2pf.gam -cpp -dpp solventLFP
gamut map: Gamuts aren't compatible
Creating smoothed near points failed
profile: Error - Failed to make perceptual gamut map transform
Hmm. Yes, that won't work, by default profile uses CIECAM02
space (Jab) internally, and the gamut needs to be compatible
with that. And CIECAM02 starts with absolute colorimetric
data, not perceptual. (I would like to fix this so that it automatically
translates between the colorspaces, but haven't thought it that
important, and in this situation you'll get the ICCV4 problem
of the result having a saturation intent bias.)
Do you mean I get a saturation-like result even when I use the
"perceptual forward gamut ;-)" of ISO Coated v2 together with a argyll
profile tailored exactly to this gamut?
What else can I do for the short term to negotiate the obstacle?
Do you think a "play nice with Heidelberg" switch for profile makes
sense in the long run?
I had a look at the colorimetric and perceptual gamuts and
they are quite a different, although the perceptual is
not the PRMG, it's a similar shape, but substantially smaller.
The source data is supposed to be on the FOGRA site
(FOGRA39L I think). You should be able to convert it
with logo2cgats and go from there.
I know it. I've build test profiles based on FOGRA39 already -- no
problem. My problem is the omnipresence of profiles like ISO Coated v2
(e.g. there is nearly the same clipping problem when using the
CoatedFOGRA39 profile from Adobe).
I'm not sure there is any way of dealing with this type of issue
without side effects. The most transparent "fix" would be to grab
the gamut from the profile supplied with the -S options from the
perceptual intent, although this would need some deep changes
(ie. I need to extract "absolute" values from the perceptual table,
to get the XYZ to feed into CIECAM02), but maybe it's not too hard.
Translating the colorspace would give another path too.
It looks like a Catch-22 situation: I can manipulate the tag table and
get a good perceptual result (and loose the colorimetric intents) -- or
I can try to make argyll's excellent perceptual gamut mapping worse, but
compatible with the majority.
At least I want to test both options before I choose one of them. And I
think I found a way to test it: in fact, fakeread doesn't offer a -i
option, but I think I can use xicclu to grab the "perceptual gamut" --
I'll try it.
Bye for now and thanks,
Klaus
Other related posts:
