[argyllcms] Re: Generating new black curve from existing profile.

Stephen Prater wrote:
One of the things I thought to try was using revfix, and specifying a new black curve - unfortunately, the profile I'm trying to convert uses XYZ as it's PCS space, so I haven't been able to try it. I think I could probably write a converter in there, but wanted to know if it would provide better results before I devoted the time to doing that. 

My experience with LUT type profiles is generally that XYZ PCS gives
poor results, simply because the XYZ space is so big (2.0^3) compared
to a typical device gamut, so the grid is used very inefficiently
used in the B2A table.

There are ways of reducing the disadvantage of XYZ PCS, by setting
up the XYZ matrix and curves in front of the LUT to better adapt the
PCS values being fed into the LUT to the size of the gamut, but
I haven't really tried this approach, since using L*a*b* PCS
generally works better.

While slightly clunky, the approach of using fake read and
re-creating the profile is a reasonable one. You can make
the test set as big as you like really (ie. 10000-100000 points
isn't unreasonable).

The other though I had was to modify profile so that it would
accept an existing profile as input, instead of a .ti3 file.
(one more idea on the list!).

Graeme Gill.

Other related posts: