Path: utzoo!attcan!uunet!dino!sharkey!itivax!umich!zip!spencer From: spencer@eecs.umich.edu (Spencer W. Thomas) Newsgroups: comp.graphics Subject: Re: SigGraph Fractal Compression Message-ID: Date: 15 Aug 89 03:16:43 GMT References: <444@mit-amt.MEDIA.MIT.EDU> <20400001@inmet> Sender: news@zippy.eecs.umich.edu Organization: University of Michigan EECS Dept Lines: 18 In-reply-to: sarrel@sioux.cis.ohio-state.edu's message of 14 Aug 89 14:56:26 GMT The technique is valid, but it is not lossless. In other words, what you get out ain't what you put in. For a noisy image (i.e., a digitized image), this may be ok. For other images (i.e., a computer generated image with an alpha channel), this is not good at all. Finally, compression is SLOW. I recall hearing a figure of 1 hour per image to compress (I could be wrong). Thus, the technique is only really useful if you are going to decompress much more frequently than you will compress. In these respects, it is very similar to another compression technique that promises extremely high compression ratios (100:1 and better): vector quantization. VQ is slow to compress, fast to decompress, and lossy. In other words, TANSTAAFL. -- =Spencer (spencer@eecs.umich.edu)