[argyllcms] Re: xicclu -g predictability issue

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Fri, 28 Jan 2011 12:17:22 +1100

Elena [service address] wrote:
Graeme, are you sure isn't there some subtle bug you never noticed ? How is that
clipped out of gamut colors cause a so big waste of time and resources ?

Because clipping properly is hard. You have to locate the closest point
on the gamut surface. I have used many acceleration techniques in the code,
and it is a great deal faster in this area than it once was. You can use memory
to reduce the size of the acceleration lists it has to look for, but
setting the ARGYLL_REV_ACC_GRID_RES_MULT environment value.

If so, then think what happens with colprof: even using -ip there are LOTS (the
majority ?) of points in the Lab/XYZ B2A table corresponding to clipped colors 
slowing
down things terribly - since the source gamut used for gamut mapping will NEVER 
cover
the entire Lab/XYZ space! Points that after all will never be used, possibly... 
If I'm
reasoning correctly, of course...

For colorimetric, the "source" gamut is the ICC L*a*b* cube. The colorimetric
B2A has to map every point in that cube to the colorimetrically closest in gamut
value.

Graeme Gill.

Other related posts: