Re: [Lcms-user] Size of pre and post linearization tables

[Kai-Uwe B. <ku...@gm...>]

The pre and post linearistation curves (or tables as called in the spec) 
for the B2A and A2B tags can contain up to 4096 segments for mft2 
(10.8) CLUT's. And for the embedded cuves (10.5) used by the v4 
lutAtoBType, I found no limitation other than the type of a uInt32Number 
for its segment number.

The more segments the finer of a control is possible. 
As well, these curves can contain asymetric numbers of segments for in/out 
or B/M/A elements.

regards
Kai-Uwe Behrmann
--
developing for colour management 
www.behrmann.name + www.oyranos.org + www.cinepaint.org


Am 06.08.07, 10:57 +0200 schrieb John Weissberg:

> We are creating BtoA tables for an output (printer profile):
> 
> Input color space: Lab
> Output color space: CMYKOVG
> 
>    Input Channels = 3
>    Output Channels = 7
>    CLUT resolution = 33
> 
> Total CLUT entries: 33 x 33 x 33 = 35937
> 
> 
> When we examine other ICC profiles, they all contain 3 extra tables:
> 
> 1. 3 x 3  XYZ to XYZ matrix that is normally set to unity matrix
> 2. Pre-processing matrix for linearizing input Lab values
> 3. Post-processing matrix for linearizing output ink values
> 
> I am completely baffled by the length requirements for tables 2 and 3.
> 
> When I examine a profile built with Profile Maker 5 (with multi-channel 
> option), it used 514 entries in both tables. Other profiles I have 
> checked use 256 entries for these tables.
> 
> What are the rules (or rules of thumb) for deciding the length of these 
> pre and post processing matrices?
> 
> I have checked the ICC4 spec and found no help there.
> 
> -- 
> John Weissberg
> PerfectProof - Accuracy Made Easy
>