Re: Absolute Colorimetry not supported with Monitor Profiles
Re: Absolute Colorimetry not supported with Monitor Profiles
- Subject: Re: Absolute Colorimetry not supported with Monitor Profiles
- From: Henrik Holmegaard <email@hidden>
- Date: Wed, 9 Jan 2002 09:12:50 +0100
Roger Breton <email@hidden> wrote:
Is this a phenomenon only applicable to LUT-type of monitor profile or is
this also applicable for Matrix/TRC-type profiles?
As I understand the question, the issue is whether matrix-based type
mntr profiles support rendering intents.
First, we have to find out what 'rendering intent' means.
According to the specification, there are two profile types classed
as 'output', type prtr printer profiles and type mntr monitor
profiles.
For type prtr profiles LUTs are required. There is no such beast as a
matrix-based ICC printer profile.
The spec requires a type prtr profile to have four 'rendering
intents' in either direction, but the spec understands 'rendering
intents' to be the entries in the profile header. This means that
what four 'rendering intents' are required, it is four entries in the
Table of Contents, as it were. The spec does not require each
'rendering intent' to reference an individual block of numbers with a
behaviour different from the other blocks of numbers. The blocks of
numbers, the LUT's, are referred to as 'tags' in the spec, and two or
more 'rendering intents' may reference the same 'tag'. Thus a type
prtr profile with the same behaviour for all 'tags' is still a legal
prtr profile, even if it is not a functional profile.
For type mntr profiles the situation is more complex, because the
spec says they can be either LUT-based or matrix-based. But at this
point things get hazy, partly due to the technical history as Adobe
and Apple came from the direction of matrix-based approaches and
others came from the direction of LUT-based approaches, and partly
due to the spec itself.
Say for matrix-based type mntr profiles, I'm not sure 'rendering
intents' in the sense of different behaviours are at all defined. For
these situations, i.e. when no behaviour is defined, relative
colorimetric behaviour is the rule.
On the other hand this does not mean that different behaviours do not
occur. Thus though in theory there should only be one behaviour with
matrix-based type mntr profiles, namely relative colorimetric, these
profiles are also able to have an absolute colorimetric behaviour.
For instance, if you select an RGB 'working space' i.e. type
matrix-based type mntr profile as source, and select absolute
colorimetric as conversion behaviour, you will get the wtpt tag
calculated into the behaviour, and the separation will in effect
proof the RGB 'working space' in the next space in your chain.
Trying to read that spec is hard work, and it can be seen as a cheese
riddled with as many and as big holes as any cheese ever was. The
holes are arranged by design or by coincidence so that all the holes
are never visible at the same time to the same person. Which ones I
overlooked in writing the above I'm sure I don't know ... -:).