[Openicc] CUPS Color Management under Linux gets into distros

[Chris Murphy]

On Feb 21, 2011, at 2:18 PM, Alastair M. Robinson wrote:
> 
> On 21/02/11 19:30, Chris Murphy wrote:
> 
> 
>> The issue is both the source space L* going to 0, compared to the destination space L* going to something more than zero,
>> and the lack of Black Point Compensation to use with RelCol. When both of these are true, Perceptual rendering should be used.
> 
> In this context BPC would be available, but I don't see any reason not to use Perceptual when converting a flat sRGB raster to a printer's profile?

It is more consistent with most ICC v2 workflows to use RelCol+BPC and does tend to produce a preferred result. The B2A0 table in output profiles can contain the same data as B2A1, or even point to it, and means Perceptual ends up being the same as RelCol.

For ICC v4 it's a different story perhaps and I might recommend Perceptual there as the default. But another possibility is to always employ BPC for either Perceptual or RelCol.


> 
>> Even if a printer is described by an RGB output device profile, the L* will not go to 0 for any real device with a properly
>> built profile. So the issue of intent is the same. Default rendering to print space needs to be either RelCol+BPC or Perceptual.
> 
> Agreed, however, when the printer's profile is RGB, Pdftoraster doesn't need to do its own conversion, so rendering intent then becomes a Poppler issue.

OK.

Chris Murphy