myk@vax5.CIT.CORNELL.EDU (03/19/89)
comp.ivideodisc people: this query was posted in comp.graphics, but i figured it would be of interest here too. In article <955@ginosko.samsung.com>, simkin@ginosko.samsung.com (Rick Simkin) writes: >I'm trying to learn about an image compression/decompression process called >DVI. This process is supposed to compress images by a factor of about 120, >and decompress them very quickly. There are two different meanings of the acronym 'DVI'. One is simply a file format used, I believe, in desktop publishing applications. I don't know whether it does significant compression or not. The other, which has no relation to the above, is Digital Video Interactive. This does achieve very significant compression. I have bad news for you, though. It is proprietary. It was developed by Sarnoff Research Labs in Princeton NJ, first under RCA, then when RCA was bought out, by GE. Intel has recently purchased this technology from GE. Its goal was to allow full-screen, full-motion video, similar in quality to ordinary (non-HQ) home VCR's, coming from a CD-ROM drive, which has a data rate of about 150K bytes per second, dreadfully small for such a goal. The way it operates is you put these special boards into your AT-compatible (the boards will be customer-shipped starting in June, and will be ENORMOUSLY expensive the first year or two). You can then do what they call 'edit-level compression', which is beastly fuzzy but allows you to develop your application software (written in C, please). In order to get 'presenta- tion level compression', you send a 1" videotape to the DVI lab, where it is compressed on a supercomputer, at a rate of three seconds of crunching per 1/30th of a second video frame, into a very compact data stream. This is recorded on a CD-ROM, and sent back to you. Those special boards contain two custom signal processors, one for displaying the on-board VRAM on a monitor, and one for pixel twiddling, especially decompressing the CD-ROM data stream. Both have 'soft' micro- code, and the commercial developers' package includes a micro- code assembler. In general, the compressed data stream contains the first full frame of a sequence, then only the CHANGES in pixels in subsequent frames. There is a book that just came out from McGraw-Hill, called 'Digital Video in the PC Environment -- Featuring DVI Technology', by Arch C. Luther (a member of the DVI development team), from McGraw- Hill, ISBN 0-07-039177-7. He talks about the capabilities of the technology, and includes some examples of C source code. Although there is a chapter on various compression techniques that one might theoretically use to compress motion video, he does not say what techniques are used for DVI. What do I think of DVI? I think it's too early to tell; MIT is working on some similarly neat things. I advise people to wait and see at this point. ---------- Mike Oltz myk@cornella.cit.cornell.UUCP (607)255-8312 Interactive Multimedia Group, Cornell Information Technologies 215 Computing and Communications Center Ithaca NY 14853