Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!usc!zaphod.mps.ohio-state.edu!mips!pacbell.com!pacbell!osc!jgk
From: jgk@osc.COM (Joe Keane)
Newsgroups: comp.graphics
Subject: Re: bit masks
Summary: Use Floyd-Steinberg with noise.
Keywords: dither
Message-ID: <3746@osc.COM>
Date: 1 Sep 90 00:35:02 GMT
References: <23900013@sunb4> <141649@sun.Eng.Sun.COM>
Reply-To: jgk@osc.COM (Joe Keane)
Organization: Object Sciences Corp., Menlo Park, CA
Lines: 14

The problem seems to be the same as dithering an arbitrary image; in this case
the image is a constant gray value.  At least, i haven't found a simpler way.

A matrix dither works OK, and is clearly easy to do.  The problem is that some
of the patterns aren't so hot.  Also, it quantizes your gray levels, so you
don't get exactly what you asked for.

I think the best way is with some sort of error-propogating algorithm.  My
favorite is bidirectional Floyd-Steinberg, which generally produces nice
patterns but has some occasional annoying artifacts.

If you don't need repeating patterns, you should add some blue noise and then
dither that.  This makes the `nicest' patterns, since they are somewhat
random, and don't have any features that stick out.