🚀 go-pugleaf

  Bill Alves and others questions about MPEG encode times and the
subsequent answers reveals one of the missunderstandings with the
standard. MPEG is really a standard for making a decoder or better a
"bitstream". In essence it is a spec for how to create a bit stream
that any decoder can play.  Within that framework there is tremendous
flexibility regarding how the bit stream is created, methods of motion
search, etc. So much flexibility is there, in fact, that one group can
lament taking "days to produce minutes" while another can brag about
1.5 seconds per frame on a 486.  I suspect that the later will have
noticable artifacts. In any event the best subjective quality imagery
will be produced only with compute intensive motion search algorithms.

 Folks will disagree on what is acceptable image quality. I come from
the consumer video side of the fence where noticable macro block edges
and quantization artifacts are not going to cut it. Video Dialtone
services at 1.2 megabits must compete with home vcr rentals and will
require a significant amount of processing - particularly for motion
search - just to make a tolerable picture. It is my opinion that image
quality considerations will limit software only and "chipset" encoders
to toy applications for the next few years at least.


-- 
|\/\/\|\/\/\|\/\/\|\/\/\|\/\/\|\/\/\|\/\/\|\/\/\/|    _________________
| (.)(| (.)(| (.)(| (.)(| (.)(| (.)(| (.)(| (.)(.)   /  Think Parallel \
C     C     C     C     C     C     C     C      _) -|       Man!      /
 Herb Taylor, David Sarnoff Research Center, email: herbt@apollo.sarnoff.com