Re: miroVIDEO DC1 tv card (fwd)
Wed, 12 Oct 1994 16:40:26 -0400

>And I'm getting more & more intrigued by the "apparent" difference in
>CU-SeeMe's compression. I understand the implementation can be/must
>be/is very different from either of the 'PEGs, but in philosophy it
>sounds remarkably similar to MPEG. Without taxing 'net-width can anyone
>enlighten me? (IE: both are panel-based, differential-update systems.)

CU-SeeMe's compression is similar to MPEG in that we assume that pixel
blocks in one frame will often be similar to those in the preceding frame.
However, rather than searching many nearby position in the preceding frame,
CU-SeeMe looks for a match only in the exact same position. Also, I
believe MPEG finds the closest match, and then transmits the difference of
actual from that prediction, whereas CU-SeeMe makes a binary decision:
either the square is close enough that the difference can be ignored or we
send the actual square (i.e., not its difference from predicted). Finally,
when it comes to transmitting an actual square, MPEG does a DCT-based lossy
spatial compression, whereas CU-SeeMe does a home-grown lossless
compression (note that CU-SeeMe can afford less loss since it is working
from the start with a 4-bit, rather than 8-bit, gray scale). Check out:
where Borre Ludvigsen has been kind enough to collect some additional
discussion of this.

Tim Dorcey
Sr. Programmer/Analyst (607) 255-5715
Advanced Technologies & Planning
CIT Network Resources
Cornell University
Ithaca, NY 14850