drtiller@uokmax.ecn.uoknor.edu (Donald Richard Tillery Jr) (11/29/90)
I can hear the hoofbeats in the background folks. The 24 bit boards are coming...the 24 bit boards are coming! There's the Toaster, Firecracker 24, HAM-E, CDTV, the A2065 (I think that's a C=one!), the Colorburst (crossed fingers) and who knows how many others.... Now for the catch (ain't there always one?)....those 24 bit RGB IFF files are HUGE! I have 11 of them right now and they range from 300K to almost 850K! Now .lzh (lharc or lz) is fine for files and fantastic for text but it just ain't good enough for these monsters (it isn't all that great for regular IFFs but I digress). What we really need is some effective IMAGE compression techniques, algorithms and eventually programs to shrink these bad boys to a manageable size.... I have heard of some such algorithms and one type in particular where the resulting image IS NOT an exact duplicate of that which started out. But the resulting reduction is up to 90% ! Now if only a .1% or less distortion is introduced, I think we can live with that (it amounts to about 1 radical pixel every other line) but I don't know any specifics about it. Rumour has it that it is the technique used for telephone face-to-face transmissions. Anyone have any info on this? Barring that, does anyone have any info on some effective image compression techniques that might make use of the 2 dimensional quality of images versus the 1 dimentional nature of programs? I think there will be a great demand for such compression in the very near future and I hope someone can put together a solution. While I'm on this subject, where can we start archiving these things. They are rather large and someone would have to have lots of storage to keep a good archive going. Any volunteers? I have some neat 24 bit ray- traces to contribute. Rick Tillery (drtiller@uokmax.ecn.uoknor.edu)
korczyns@dtoa3.dt.navy.mil (Korczynski) (11/29/90)
Rich, What about JPEG compression? Adobe's PhotoShop on the MAC gets some good results using this technique. There has been a couple messages concerning JPEG compression in Comp.Sys.Graphics. Hopefully someone will develop/port this to the Amiga when 24bit becomes more viable. Also C-Cube has introduced hardware which does image compression/decompression in real time. This chip may make 24 bit real time animations possible. Joe Korczynski Code 2723 David Taylor Research Center Annapolis, MD 21402
hill@evax.arl.utexas.edu (Adam Hill) (12/01/90)
But the SW implementation of the JPEG compression algorithm is freely availible. (At least thats what BYTE said). I think its main use will be in REALLY lots of COLOR cards. Which dont exist yet for the Amiga. I'm talking about 32 bit (24 color, 8 z-buff,alpha, whatever). The process seems simple though... :-) The image looks faded out, like when you leave a picture out in the sun. One of the interesting things in the article was the statement about JPEG and full motion video: ".... Second, even the blazing speed of the 30MHz CL550 is only adequate for real time video compression using hard drives with transfer rates of 1MBps or more." Doesn't sound like a problem for Amiga HD controller. (At least for the Hardframe 1.7MBytes/sec). I wish Visoin Quest luck. -- adam hill hill@evax.arl.utexas.edu Make Up Your Own Mind.. AMIGA! Amiga... Multimedia NOW Most Common Phrase at DevCon '90 - "Shhhhhhh.."
mark@calvin..westford.ccur.com (Mark Thompson) (12/01/90)
In article <1990Nov28.192940.30995@uokmax.ecn.uoknor.edu> drtiller@uokmax.ecn.uoknor.edu (Donald Richard Tillery Jr) writes: >What we really need is some effective IMAGE compression techniques, >I have heard of some such algorithms and one type in particular where >the resulting image IS NOT an exact duplicate of that which started >out. But the resulting reduction is up to 90% ! Now if only a .1% >or less distortion is introduced, I think we can live with that (it >amounts to about 1 radical pixel every other line) but I don't know >any specifics about it. JPEG image compression which has been discussed extensively lately in some of the other newsgoups and can achieve compression ratios up to 200:1. However, it is a lossy compression technique which will degrade the image. For video images, discernable loss begins at about a 20:1 compression ratio (note that the amount of compression performed is user variable). But remember that video images have half the bandwidth in color as they do in intensity and intensity is limited to 4.5Mhz. What this all means is that while this degree of compression is barely noticeable on a video image, it will be BUTT-UGLY on your gorgeous crisp RGB image. It is left to be seen how well this compression will fair will a clean RGB image. Also of note, the compression time can be rather lengthy. For those interested, the JPEG algorithm uses a discrete cosine transform followed by quantization (where the loss is incurred) and Huffman encoding. +--------------------------------------------------------------------------+ | Mark Thompson | | mark@westford.ccur.com | | ...!{decvax,uunet}!masscomp!mark Designing high performance graphics | | (508)392-2480 engines today for a better tomorrow. | +------------------------------------------------------------------------- +
bbs00010@uafcseg.uucp (Don Kennedy) (12/02/90)
As you may or may not know, we are finishing up our 16 bit stereo audio sampler. I am truly amazed at what we can do with our Amys. Last night I watched while Greg, our engineer, was sampling some Joe Satriani from a CD into his old Seagate 65ms drive, while perfmon was watching CPU usage. Mind you, we were using Greg's 2091 (DMA), but during sampling and playback I never saw the CPU time run over about 70-75% ON AN UNACCELERATED 2000. Playback ran more like 60-65%. Fast memory is required, but only 2 64k buffers. The Eagles' "Seven Bridges Road" takes up about 29meg (2 min. 50 sec. sampling time...). Don Kennedy Vision Quest
kurt@tc.fluke.COM (Kurt Guntheroth) (12/07/90)
Most of the nifty new image compression algorithms suffer from one or more of the following problems * very time consuming to encode (fractal based algorithms) * require special hardware to do rapidly * patented (now that you can patent algorithms...) This will make it hard on us poor hackers