Re: some thoughts on CIELAB
Re: some thoughts on CIELAB
- Subject: Re: some thoughts on CIELAB
- From: "Bruce J. Lindbloom" <email@hidden>
- Date: Mon, 10 Feb 2003 10:02:40 -0600
Roberto Michelena wrote:
>
I guess it would take a really complex CMM to do its interpolation in any
>
other space than Lab or XYZ, given that the table is handed to it in one of
>
those spaces. And by doing interpolation in a space which isn't perceptually
>
uniform (in which distances mean different things depending on where you
>
are!), and moreover whose perceptual curvature can reach higher orders than
>
the interpolation's order, and whose deformities are different for different
>
parts of the color space...
As the number of grid points in the profile increases, the color space in
which the interpolation occurs, and the interpolation method used, become
less important. The discussion about Lab's "curved" hue lines and the
problems resulting from them has much more to do with the process of
creating the profile than it has to do with using the profile (by the CMM).
Regarding the best space for interpolation (Lab vs. XYZ), that depends on
the nature of the device being profiled. Some devices (like monitors) are
better modeled in XYZ than Lab. For other devices (like printers), the
reverse is generally true. Also note that the "lut8Type" and "lut16Type" tag
types used by A2Bn and B2An tags have input LUTs that allow a more
perceptual distribution of grid points, even for XYZ grids.
Roberto again:
>
If such a CMM were to exist that makes use of a third space for
>
interpolation, still the table grid is defined in either Lab or XYZ, leading
>
to uneven sample density (and possible holes) in such third space.
I believe the current ICC specification allows plenty of room for the clever
CMM author to do interpolation in whatever secret color space he/she wants.
If there is continuity in the Lab or XYZ grid of the profile (as there
should be), there would (should) also be continuity in the secret color
space.
I would again like to emphasize that the PCS of ICC profiles is a color
space for *communicating* and *exchanging* color information. Profile
creation programs and CMMs are free to use whatever color spaces they want
for *computing* color information. The ICC specification does not prohibit
this. Profiling packages and CMMs often (usually? always?) use Lab or XYZ
for *computing*, for reasons of convenience, ignorance or lack of better
choices, but don't blame the ICC for problems resulting from this.
--
Bruce J. Lindbloom
email@hidden
http://www.brucelindbloom.com
_______________________________________________
colorsync-users mailing list | email@hidden
Help/Unsubscribe/Archives:
http://www.lists.apple.com/mailman/listinfo/colorsync-users
Do not post admin requests to the list. They will be ignored.