Re: Whitepoint problems
Re: Whitepoint problems
- Subject: Re: Whitepoint problems
- From: Jan-Peter Homann <email@hidden>
- Date: Sat, 21 Apr 2001 18:50:47 +0200
>
Discussion in the colorsync Mailing list:
Dan Sutton wrote:
>
>
My concern is 1) should(could?) the white
>
point be off in a CMM (esp. One that has been used thoroughly in the past)
>
and 2) I have heard the Apple CMM is very similar to Kodak's but maybe I
>
should "shop around" to see if there is a better option out there, or if it
>
even makes that much of a difference. For now, I'm sticking with Apple until
>
I see something I don't like about it.
Understanding the whitepoint-problems in colormanagement is complex like
understanding PostScript-Colormanagement or adjusting an offsetpress for printing
IT-8 charts.
Whitepoint-problems are always resulting from the combination of the icc-profiles
and CMMs. There is no clear declarition in the ICC-specs to proof an ICC-profile
or an CMM for correct handling of the whitepoint !!!!
It is not only important which CMM is combined with which ICC-profile, it is also
very important wich version of the CMM is used and which version was the software
for profile generation. Also it is important, if an ICC-profile has been edited
with another software.
Whitepoint problems are well known in two fields:
- separation from RGB- or Lab-Files
- Proofing with the relative colorimetric intent (no paperwhite)
In both cases white should stay white, and is does not. What is happen internaly
?
Necessary to know first is the concept of the gridpoints. Every ICC-profile has
internal tables (Luts) . This Luts are connecting the RGB or CMYK values for a
given rendering intent with Lab values. As it is impossible to have complete
entries in a LUT for all 16.670.000 CMYK-Colors, there are only some thousand
entries in a LUT. The gridpoints are describing which colors are real stored in a
LUT. A ICC-profile with 3x3x3 gridpoints has 27 entries in a LUT. An ICC-profile
with 32x32x32 gridpoints has 32.768 entries.
But not only the ICC-profile has gridpoints. Also the CMM is using an internal
grid for colorcalculation. The best calculation is done when the CMM and the
ICC-profile has the same number of gridpoints. Eg. 32x32x32.
ICC-Profiles from PrintOpen 3.0 and the actual Heidelberg CMM in ColorSync and
ICM are using a 32x32x32 grid. But a 32x32x32 grid has the problem, that there is
no direct entry for White either in the ICC-profile or in the CMM. The Whitepoint
is always calculated by the nearby entries in the grid. Profiles and CMMs with
33x33x33 gridpoint are handling the whitepoint much better. E.g. the actual
Apple CMM or profiles from other vendors. Mathematically very interesting is the
comination of ICC-profiles with 32 gridpoint with an 33 gridpoint CMM or vive
versa....
Till PrintOpen 3 and the actual CMMs Heidelberg has preferred 32x32x32 gridpoint.
With PrintOpen 4 and the upcoming CMMs Heidelberg is going to 33x33x33
gridpoints.
But the number of gridpoints is only one question. Another thing is, the method
in the CMM of using the nearby-values in the grid for color-calculation.
The one and only solution I see, is a reference-implemtations of a CMM and a set
of reference-profiles by the ICC itself.
Every developer (and user) of a profiling application would be able to check the
correct handling of the whitepoint in his profiles against the ICC-reference CMM.
Every developer (and user) of a new CMM would be able to check his CMM against
the set of reference-profiles from the ICC.
So all colormanagement-users:
If you don4t want have whitepoint-problems anymore, vote for a reference CMM and
reference-profiles at the ICC User feedback taskforce.
for europe:
mailto:email@hidden
for US:
mailto:email@hidden
--
********** neue Privat/Rechnungsadresse ************
Jan-Peter Homann
mailto:email@hidden
http://www.colormanagement.de
Privat/Rechnungsadresse: B|ro/Lieferadresse
Jan-Peter Homann Jan-Peter Homann
Richard-Sorge-Str. 63 c/o Appel-Grafik
10249 Berlin Zossener Str. 55
Germany 10961 Berlin
Germany
fon/fax: 0049 30 440 89 72 fon 0049 30 28 306-290
fax 0049 30 28 306-209
leo 0049 30 28 097 709
********** neue Privat/Rechnungsadresse ************