[comp.graphics] JPEG as a route to affordable computer animation

woolstar@nntp-server.caltech.edu (John D. Woolverton) (11/15/90)

I just returned from Comdex, where one of the principle things
I went looking for was C-Cube.  C-cube had a nice booth with
companies showing off both a Macintosh and IBM version of video
boards that could do real-time playback off of hard drives.
In the center of the booth was a Mac playing back the ending of
Star Wars and other video snippits at 25:1 compression from
some nameless hard drive, and other displays were loading large
images with little compression at nice speeds off of slow
disks (like 100-200k/sec).

   Now at 25:1 compression, you can tell the difference between
original and JPEG on a RGB monitor, but it is very difficult on
a TV.  A demonstration tape from C-Cube itself was showing off
a very colorful moving image at 16:1, with even better results.
I was curious if one could find a disk drive that could spit
out enough bytes to keep out, and found several.  Both Segate
and HP, have 5,1/2" drives in 600, 1200, and 1600 MB capacities
that will support > 3MB/s sync SCSI transfers.  The high end
HP claims were around 5MB/s.  And if you are insane, you can
get the 8" SABER IPI2 drives, with 3 GB capacity and unreal transfer
rates.  (Silicon Graphics has a demo where they display 
16bit 500x300 real time raw off of two IPI2 drives.)
   
   As far as how this lossy compression compares against
the $100k professional machines, I have a lot of experience working
with the Abekas, a digital frame recorder.  It too did bandwidth
sampling and threw away about 33% of the incoming data, storing
YUV with U and V averaged across pairs of pixels.  You could
tell the difference between the Abekas and a RGB original as well.
Again, once you went to NTSC, nobody could tell the difference.

   So, I'm looking to jump on the JPEG bandwagon, dump the single
frame recorder, and video controllers (about a $12k investment),
and work with a much simpler system.

	John D. Woolverton, video engineer
	   woolstar@cobalt.caltech.edu

	   "Read my MIPS: No new VAXes"

mcdonald@aries.scs.uiuc.edu (Doug McDonald) (11/16/90)

In article <1990Nov15.023216.12172@nntp-server.caltech.edu> woolstar@nntp-server.caltech.edu (John D. Woolverton) writes:
>
>   Now at 25:1 compression, you can tell the difference between
>original and JPEG on a RGB monitor, but it is very difficult on
>a TV.
On some pictures.

>A demonstration tape from C-Cube itself
Meaningless. Lets see demonstrations tapes from neutral parties.
Or people out to find problems.

>
>   So, I'm looking to jump on the JPEG bandwagon, dump the single
>frame recorder, and video controllers (about a $12k investment),
>and work with a much simpler system.
>

For broadcast TV, quality is not very important, NTSC is so crappy
anyway. Lets think about real high definition TV, not poorer than
NTSC. 

Doug MCDonald