gfs@abvax.UUCP (12/04/87)
I have a question for the more knowledgable about the Chaotic Compression technique. I read the article in Computer Graphics world a while ago and have been rolling around the idea in my head. It seems to me that since the algorithm is inherently finding a semi-random 'fit' for the features of the picture with pieces of fractals, a small change in the scene (as in a sequence of animated frames of something moving slowly) would give rise to larger random detail changes in the frame to frame compressed/decompressed scene. My argument is very analogous to the problem of animating a scene while using a color pallette. If you allow each individual frame to decide on its 'best representation set' of pallette colors, you will get 'color flickering' between scenes that is very noticable and bothersome to the eye. The solution to the color pallette problem is to choose a static color pallette set that best represents the colors in the whole scene. The pallette can be changed at moments of large visual discontinuity, i.e. scene cuts or shifts. I predict that if someone made an animated film using the Chaotic Compression algorithm, there would be a large amount of 'detail motion' kind of like watching a very very grainy old film. The edges of areas moving slowly would squirm and jump about in an almost Brownian motion. Am I right? Does anyone have further insights? Greg Shay Allen Bradley Co.