Alastair M. Robinson wrote:
As far as I know, Gutenprint supports this printer in 6-colour mode now (and for the record, I've successfully built it as a DLL on Win32 before now, with only very minor changes) - and Robert Krawitz has put a *lot* of effort into the channel splitting and n-colour generation in recent months - so there's useful reference code there if nothing else :)
I know about Gutenprint, but it is not close to the approach I want to use. As a general rule, equation driven separations work badly. There's no linearisation or calibration system, and no ICC profile/device link/separation device link data flow. I'd like to set it up to emulate other inksets too, for testing profilng code against. The main reason for using the printer is as a platform for developing general N color separations. I've also got a nice stocastic screen I'd prefer to use, and some tricks with multi-level dots that Gutenprint may not be doing.
I'm starting to see how limited the ICC format is, and why some people deprecate the perceptual intent.
Yes, the inclusion of the gamut mapping into the B2A processing is fundamentally flawed, since it depends on the source gamut, which is not actually connected with the output profile. There has been some move to address this issue in ICC V4, but it still doesn't cover it, since there is no defined PCS gamut (yet). Even with this in place, I have my doubts about the quality loss in concatenating two gamut mappings (input -> PCS, PCS -> output).
You get a much better result from all points of view, if you are doing lots of conversions between two colorspaces, if you use a device link profile (icclink -G etc.).
Graeme Gill.
