Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Posting-Version: version B 2.10.2 9/18/84; site Navajo.ARPA Path: utzoo!watmath!clyde!bonnie!akgua!whuxlm!harpo!decvax!decwrl!Glacier!Navajo!fournier From: fournier@Navajo.ARPA Newsgroups: net.graphics Subject: Re: fractal mountains Message-ID: <161@Navajo.ARPA> Date: Wed, 26-Jun-85 18:34:10 EDT Article-I.D.: Navajo.161 Posted: Wed Jun 26 18:34:10 1985 Date-Received: Sat, 29-Jun-85 01:43:17 EDT References: <241@kovacs.UUCP> Organization: Stanford University Lines: 25 *** REPLACE THIS LINE WITH YOUR FRACTAL *** Indeed the triangular subdivision you mention is better than the earlier one. It still has a big drawback though, in that the subtriangles created will be recursively subdivided without any correlation with the adjoining subtriangles. If you want to approximate fractional Brownian motion (one of the random fractal processes), this is incorrect since the 2-D version is non-Markovian, meaning that one side of the surface cannot be computed knowing only the boundary. A better subdivision scheme can be found in the Fournier/Fussell/Carpenter paper in CACM, June 1982, or the Piper/Fournier Siggraph paper of 1984. Of course a rejoinder by B. Mandelbrot and a reply can be found in the following CACM issue of August (or July). Several people are currently working on new schemes to generate 2-D approximations of fBm. There will be a tutorial at Siggraph on fractals, with Loren Carpenter and Benoit Mandelbrot together again for the first time. This should be fun! If anybody is interested in 3-D fractal (or approximations thereof), I can If anybody is interested in C code to generate 3-D fractal (and the code is short indeed, but no display is included), mail to me. It will take a few days, since I just got here and I have to bring most everything up anew. Remember that 3-D here means the process generate a volume, not a 2-D surface to be inbedded in a 3-D space.