[comp.graphics] Video compression and realtime decompression techniques

sean@cadre.dsl.PITTSBURGH.EDU (Sean McLinden) (06/13/89)

My apologies for posting this as I remember that someone posted
a blurb on it awhile back. I've searched my ACM and IEEE journals
back a year or two and have not found it. Anyway...

Can anyone provide me with pointers to references on realtime
digital video decompression and video compression using super-
computer techniques?

E-mail responses would be appreciated. Please don't assume that
someone else will answer this one. Thanks in advance.

Sean McLinden
Decision Systems Laboratory
University of Pittsburgh

hutch@celerity.uucp (Jim Hutchison) (06/15/89)

sean@cadre.dsl.PITTSBURGH.EDU (Sean McLinden) writes:
>My apologies for posting this as I remember that someone posted
>a blurb on it awhile back.

No need to apologize, this is a relatively current topic.  I believe
the reference I saw was a short piece by IEEE with a large hot air balloon
image featured in it. (Well behaved image, few colors)

>Can anyone provide me with pointers to references on realtime
>digital video decompression and video compression using super-
>computer techniques?

The algorithm I read was a combination of quadric subdivision and a
modified fft.  The algorithm was really quite witty, my apologies that
I can not provide you with a direct reference.

Here is the gist of it:

	1) encode the grey component (I), I'm real fuzzy on the algorithm
	   they were using for this.  It is the highest bandwidth channel
	   since it is what people are most sensitive to.

	2) Y & Q are compressed using an fft algorithm which uses changes
	   in angle instead of magnitude.

	Note:  There is data loss.

This got them down to 1.2 (I recall) bits per pixel in real time (video
rate).  Note that the bits/pixel relates directly to the sampling rate
since video is an analog signal with "unlimited" horizontal positioning
(positioning not resolution).

This task could certainly be done with a processor board sporting some video
ram and a DSP chip, no need for a super computer.  A TAC or a Pixar might even
be able to do it.
-
/*    Jim Hutchison   		{dcdwest,ucbvax}!ucsd!celerity!hutch  */
/*    Disclaimor:  I am not an official spokesman for FPS computing   */