Path: utzoo!attcan!uunet!lll-winken!lll-lcc!pyramid!octopus!pete From: pete@octopus.UUCP (Pete Holzmann) Newsgroups: comp.graphics Subject: Re: Fractal Compression Summary: See article in CGT. There's a lot of hype goin' around! Keywords: fractals, compression Message-ID: <228@octopus.UUCP> Date: 17 May 88 04:11:09 GMT References: <686@thalia.rice.edu> Reply-To: pete@octopus.UUCP (Pete Holzmann) Organization: Octopus Enterprises, Cupertino CA Lines: 57 In article <686@thalia.rice.edu> wasg@rice.edu (Eric Salituro) writes: >I fail to see what the hoopla over fractal-based image compression is all >about. In the last two weeks, I've read a couple of breathless articles >proclaiming 1000 to 1 reduction but all I've seen as evidence are a couple >of low-res pictures. >I don't know what the originals looked like, but I think we could be wowed >more effectively after seeing a comparison. I think there is a forest image >running around that doesn't look as good as Reeves' particle-based forest >from "Andre and Wally B." >Both articles claim that after the compression process "the original can be >thrown away." Isn't that a little hyperbolic? From what I understand The article in Computer Graphics Today (April '88) is good reading. Salient points gleaned from it: - High compression ratios come from using very high resolution pictures, presumably with little information content in the extra resolution. At 1Kx1Kx24bits, they claim frequent reductions of 3000:1 (although their description seems to assume that the color table is not contained within the final compressed image). At 2Kx2Kx24bits of the same picture, they can supposedly get to 10K:1 - The compression process is slow: it takes human-interaction intelligence to find the compressible algorithms, on the order of 1000 hours per picture. That is no typo! They think that a PC based version that will do the compressions in only 1 hour. Of course, the hardware involved will cost $500,000. (Again, not a typo). This is a 'PC'? Would that I had a few!!! - Decompression takes 30 minutes now on a fast box. They are hoping to eventually hit 20 seconds (on future parallel processor technology), and someday in the distant future to get 30 fps. - The example shown DOES look rather hokey. It is pretty good for 'computer generated art', but nothing like real life! I can't believe they think this is anywhere NEAR as good as the original image. YUCK! Since when have you seen a mountainscape where there are only 2-3 kinds of tree shapes, and most branches look identical, and there are maybe 6 shades of green in the whole forest, and...... Let's see.... take a 2K by 2K by 24 bit image. That's 12 MB. Reduce it to 400 x 256 by 6 bits, with a 64 x 24 LUT (thats 75K now). Use RLL encoding for initial compression (oh... 30K I guess). Run it through LZW compression to get it down to 24K. I claim my method goes from 12MB to 24K! Amazing. 500:1 compression, *FAST* reconstruction (just dither to get more resolution, it's easy!), and you can throw away the original! Can I have a big government grant too??? (:-) :-) :-) I'm sorry, I just couldn't resist!) I realize this research may be good basic science, and it may lead to useful results sometime in the distant future... BUT THESE GUYS ARE MILKING IT FOR ALL THE P.R. THEY CAN GET. I am not impressed. -- OOO __| ___ Peter Holzmann, Octopus Enterprises OOOOOOO___/ _______ USPS: 19611 La Mar Court, Cupertino, CA 95014 OOOOO \___/ UUCP: {hpda,pyramid}!octopus!pete ___| \_____ Phone: 408/996-7746