Re: [Gimp-print-devel] Epson Stylus 660

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

   Date: Mon, 20 Mar 2000 19:07:01 +0100
   From: Thomas Tonino <tt...@bi...>

   It might be worthwhile to try the few matrices that are on my website (
   http://people.a2000.nl/ztonino/dither/files/64C for example) as the
   randomizing data in the FS process. It is bound to be less visible than
   true random noise. If the patent problems are too complicated for a
   production environment, I'm willing to spend a day or more on an
   "artistically created" equivalent. The good thing about void-and-cluster
   generated random data is that it is of higher frequency than normal
   random data: blue noise as opposed to white noise.

The current dither code isn't general enough to be able to handle
arbitrary matrices.  It's probably worth generalizing it some time.
Right now it uses an adaptive matrix (or what I call an adaptive
matrix, at any rate :-) ).