Re: follow up results on "Lab in always just right"
Re: follow up results on "Lab in always just right"
- Subject: Re: follow up results on "Lab in always just right"
- From: "Bruce J. Lindbloom" <email@hidden>
- Date: Fri, 14 Sep 2001 10:55:34 -0500
Tom Vanderlinden wrote:
>
After some off list communications, exchange of pictures
>
and testing between Steve Upton of CHROMiX and myself,
>
we have found that the 3-D representation of the Lab space
>
as rendered by ColorThink is faulty when ColorThink uses
>
the Apple CMM, which is controlled by what CMM you select
>
in the ColorSync control panel.
I have three things to say about this.
1) Many of the RGB reference space definitions are based on non-D50
reference whites. Notable examples include Adobe (1998), sRGB, BRUCE RGB
and others. In order to cast these profiles into ICC conformance (which is
D50 based), the reference primaries must be "adapted" from their original
reference white to D50 when the profile is _made_. There's more than one
way to do this and different methods will place the primaries in different
locations in XYZ (or Lab) space, producing gamuts of differing sizes and
shapes. There is no provision in the ICC format for the profile creator to
indicate the adaptation method used during profile creation.
Later on, when the profile is _used_ (by a CMM), these primaries may need to
be further adapted to a non-D50 reference. Since the CMM cannot determine
by examining the ICC profile contents which method of adaptation was used
during creation, its own adaptation algorithm cannot be matched to it with
certainty. This may be the reason why Tom has found different results with
different CMMs -- each may use its own adaptation method.
2) There is a rather loose use of the term "Lab space" in this discussion
regarding whether or not a gamut lies within or without. It appears as
though what is referred to is actually the Lab encoding space used by the
ICC profile format, namely any coordinate whose L* is in the range [0, 100]
and whose a* and b* are in the range [-128, 127]. This is an _encoding_
limitation, not an Lab limitation. Most (about 65%) of the possible
coordinates allowed by by this encoding do not even represent real colors.
3) Which leads to the display of the gamut of a monitor (or RGB working
space) profile in Lab (note: monitor ICC profiles are XYZ based, not Lab
based). Since the reference primaries do not need to represent real colors
(and many of the reference color spaces have unreal primaries, either by
design or by way of adaptation which forces them to be unreal), a gamut
viewer must do something about conversion of illegal XYZ values to Lab for
display. For example one of the reference primaries may have a negative X
or Z value. The CIE standard does not specify how to perform XYZ to Lab
conversions in such illegal cases. This might be a problem for a gamut
viewer who must perform this conversion "seat-of-the-pants", without
guidance of a standard.
--
Bruce J. Lindbloom, Pictographics Intl. Corp.