Re: [Gimp-print-devel] New dither matrix

[Robert L K. <rl...@al...>]

It looks promising.

One problem, though, is that the same matrix is used for everything.
We really need eight uncorrelated (or perhaps two pairs of
anticorrelated) matrices to do it right for 6-color or variable dot
size printers, though.

For the four colors, we could use two pairs of anti-correlated
matrices: probably one pair for K and Y, the other pair for C and M.
If there are slight correlations, it may not matter too much.

However, for variable dot sizes or 6 colors, we have another problem:
we need to pick which color to use.  That really does need to be
uncorrelated with the decision about whether to print at all,
otherwise there will be blotches or odd transition regions.

   Date: Wed, 17 May 2000 22:06:57 +0200
   From: Thomas Tonino <tt...@bi...>

   Also, Random FS (in the patched version) seems to have some artefacts.
   This time I printed a pattern of colored blocks. Random FS shows two of
   these with aright hand edge that gets slighly darker down the page, and
   there is a block (which is a dirty yellow) where 4 magenta and a cyan
   pixel 'escaped' at the bottom. This may be the 'random' in random FS
   though.

The "random" simply means that a random number is used to perturb the
threshold value, as opposed to a matrix value for "hybrid" FS.

   Finally, FS looks better on photographs because it has a sharpening
   effect that is missing in ordered dithers.

Sharpening?  I would think precisely the opposite.

   I take the conclusion that FS is better for dark tints while ordered
   dither now is better for light tints.

That's the theory behind the adaptive dithers.  Since this matrix is a
lot better we can probably try pushing up the cutover point to reduce
the error diffusion artifacts.

-- 
Robert Krawitz <rl...@al...>      http://www.tiac.net/users/rlk/

Tall Clubs International  --  http://www.tall.org/ or 1-888-IM-TALL-2
Member of the League for Programming Freedom -- mail lp...@uu...
Project lead for The Gimp Print --  http://gimp-print.sourceforge.net

"Linux doesn't dictate how I work, I dictate how Linux works."
--Eric Crampton