Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!usc!cs.utexas.edu!uunet!ns-mx!pyrite.cs.uiowa.edu From: jones@pyrite.cs.uiowa.edu (Douglas W. Jones,201H MLH,3193350740,3193382879) Newsgroups: comp.compression Subject: Re: looking for info on image compression Message-ID: <5043@ns-mx.uiowa.edu> Date: 25 Mar 91 19:48:07 GMT References: <5040@ns-mx.uiowa.edu> Sender: news@ns-mx.uiowa.edu Lines: 38 From article <5040@ns-mx.uiowa.edu>, by drenze@umaxc.weeg.uiowa.edu (Douglas Renze): > ... a set of algorithms for GIF image compression. I haven't looked at > them, but I would assume that they're fairly complete & that they could > (maybe) be adapted to other image types ... > GIF images use Ziv-Lempel compression, which is not a particularly good way to compress images. The problem is, it operates on strings of successive letters, and it works best if short strings of letters have some significance and are likely to be repeated. In images, repetitions of any string, when they occur, are usually accidental, and strings of pixels, reading left to right, are lousy predictors of the next pixel compared to 2-dimensional neighborhoods. I did an experiment using my splay-tree based compression algorithm with just one state in its source model, and it equalled the performance of Ziv-Lempel compression on digitized portraits. Multi-state compression models for images can do much better (and my splay-tree based algorithm is not intended to be a space-optimal compression algorithm). These results were published in my CACM paper of Aug, 1988. Further results of mine concerning use of splay-tree based compression on images will appear in the 1991 Data Compression Conference two weeks from now. (PS: Does the use of Ziv-Lempel compression for GIF images mean that the question of patent rights for the Ziv-Lempel algorithm applies to GIF images? My source code for one GIF viewer had a note saying that the uncompression code was borrowed from the UNIX compress command, and I understand that there are patent infringement questions currently being asked about UNIX compress.) (PPS: I give away copies of my splay-tree compression code to anyone who asks for it.) Doug Jones jones@cs.uiowa.edu