Re: mapping extreme colours perceptually
Re: mapping extreme colours perceptually
- Subject: Re: mapping extreme colours perceptually
- From: Thomas Knoll <email@hidden>
- Date: Tue, 21 Aug 2001 07:57:15 -0700
Neil,
You seem to have discovered how much of a kludge "perceptual"
rendering really is with the current ICC profile model.
The standard description of perceptual intent: "smoothly compresses
the entire gamut of the source profile into the gamut output profile
without clipping" is absolute b***s***. It does not work this way.
It is impossible to make it work this way with the current ICC
profile model.
The basic problem is that source profile and the destination profile
are completely independent. All the source profile does is tell the
CMM how to convert source device colors to the PCS space (basically
Lab space). The source profile has no idea what the destination
profile is going to be. All the destination profile does is tell the
CMM how to convert colors in the PCS space to device colors. The
destination profile has no idea what the source profile was.
So how is perceptual intent *really* implemented? Answer: the author
of the profile building tool makes some wild guess of a "typical"
source gamut, and builds some fixed desaturation and darkening (of
high saturation colors) into the destination profile. Because the
author wants the resulting profiles to work well for a wide range of
source images (including images already mostly in gamut), the fixed
gamut compression is usually limited to only minor amounts. In any
case, *huge* amounts of the PCS space are still clipped by the
destination profile.
And no, the CMM does not "handle it". I don't know of any CMMs on
the market that perform any smart gamut mapping. All of them just
use the look up tables built into profiles to convert from source to
PCS and from PCS to destination. The only non-standard CMM
processing in common use the Adobe's "black point compensation",
which is more tone mapping rather than gamut mapping.
Thomas Knoll.