[comp.protocols.tcp-ip] Compressed Video Transport Requirements Summary

kwe@BU-IT.BU.EDU (Kent England) (02/27/91)

	I didn't get an answer to my question, but I did get some
info on the new ISO standards for compression of still frames and
motion pictures (correlated frame sequences, actually).  Herewith
is what I got, including news on a satellite broadcast system that
is going to use compression to get more channels to the end-user.

	You might want to check out comp.ivideodisc, comp.multimedia,
comp.graphics, rec.video.satellite for more info.

	--Kent

_________________________

From c3!c3.PLA.CA.US!rww@fernwood.mpk.ca.us Wed Feb 13 05:43:16 1991
Date: Wed, 13 Feb 91 02:26:55 PST
From: rww@c3.PLA.CA.US (Richard W. Webb)
To: kwe@bu.edu, barmar@think.com
Subject: Video Compression
Newsgroups: comp.protocols.tcp-ip
Organization: C-Cube Microsystems, San Jose, CA

I have a keen interest in the use of video compression in
a networked environment.  I have collected a number of
articles describing what is available.  I have edited it,
so be forewarned that this is a somewhat biased viewpoint.

JPEG is a Still-image compression technique.  Video is generated
by rapidly displaying consecutive independent frames.

MPEG is a Moving Picture approach that utilizes some of the
similarities that exist between consecutive frames.  It also
includes sound compression and a number of other "system"
features.

Both of these approaches have problems, but, in some sense,
they represent a "optimal" tradeoff between complexity and
quality.  They are also highly parameterized and allow a
broad adjustment in capabilities.

One final note on bandwidth requirements.  Both standards make
allowances for fixed-rate transmission and reception.  This
fixed rate can be updated periodically (at the encoder end)
based, say, on an extimate of network congestion.  The update
rate can be sent, at most, once per frame (60 or 50 frames per
second), but once per 6 or 10 frames is more typical.


=======================================================================

From: cnh5730@calvin.tamu.edu (Chuck Herrick)
Date: 8 Feb 91 16:57:29 GMT
Organization: Geodynamics Research Institute, Texas A&M University
Lines: 36

JPEG stands for Joint Photographers Engineering Group or some such,
and this is a "standardizing body" which has formalized a 2-d (thus
useful for images) compression/decompression scheme which uses a
modified cosine transform to "do its thing." Right now, the JPEG 
compression sceme is "cutting edge" in the image processing game.

Please note that the JPEG compression is what the folks in the trade
call "lossy"... this means that you may lose at least some information
in the compression process (you might be pretty impressed at the fact
that in many cases, you can't see the difference between an original
image and its compressed-decompressed sibling).

JPEG is neither strictly hardware nor strictly software. A company
called C-Cube is in the process of trying to implement the JPEG
modified cosine transform compression/decompression algorithm in
a chip (i.e. in hardware). When, and if, they succeed, the C-Cube
chip will allow real-time compression, and will be great for real-time
video grabbing. However, the chip is still in process, and until 
someone gets a bug-free JPEG chip to the market, one will need to
implement the JPEG algorithm in software. 


=======================================================================

>From: jim@newmedia.UUCP (Jim Beveridge)
Newsgroups: comp.ivideodisc,comp.multimedia
Subject: Re: DVI questions
Date: 15 Jan 91 15:09:30 GMT
Organization: New Media Graphics, Billerica, MA

The first chip to do JPEG is from C-Cube, and they are currently only
shipping the still frame version of the chip.  The real time version
is still not available.  Even in compressed form, the bandwidth
required for a full JPEG screen far exceeds the abilities of an
IBM bus to transfer.  (I don't believe it to be a problem for the
Apple NuBus)  JPEG still requires LOTS of data moving around.
To keep track of it, you pretty much require the full resources
of the system to move it off the hard disk and pump it into the
chip fast enough.

Of course, there are ways around this problem with a private
bus and private hard drives, but that is $$$.

The MPEG standard is still under discussion and won't be ready
for at least a year.  Don't expect commercially available MPEG
boards for a couple of years.

DVI is shipping now, but is VERY expensive, particularly for
the production level video that requires that you send a tape
to Intel.  The "home-brew" comperssion that the DVI chips
now do is very grainy and not suitable for production.
The good news is that the production level does not require
almost the entire power of the CPU to keep the picture running.

		Jim

=======================================================================

From: jandreas@pro-graphics.cts.com (Jason Andreas)
Newsgroups: comp.graphics
Subject: Re: JPEG in VLSI
Date: 26 Jan 91 07:56:21 GMT

  C-Cube MicroSystems
  399-A West Trimble Road
  San Jose, CA 95131
  voice. (408) 944-6300
    fax. (408) 944-6314j
  
   Ask for Jill Milton or Clint Chao.

======================================================================

From: ltran@pluton.matrox.com (Linh TRAN)
Newsgroups: comp.graphics
Subject: Re: Animation Standards JPEG
Date: 30 Jan 91 21:49:37 GMT
Organization: Matrox Ltd.

I think MPEG is working on standard for motion picture (animation and sound
included). The MPEG uses fundamentally the same algorithm regarding compression
scheme (it use Discrete Cosine Transform, follow by Quantization of coefficients
and Huffman encoding, for intra frame compression). JPEG mandate is to arrive
at photographic resolution compression.                

======================================================================

	Yakov Rekhter of IBM was kind enough to forward this on to me.
-kwe 

____________________________________________

From YAKOV%YKTVMZ@IBM.COM Thu Feb 21 08:23:57 1991
Date: Thu, 21 Feb 91 08:19:27 EST
From: YAKOV@IBM.COM
To: bu-it.bu.edu!kwe@uunet.UU.NET
Subject: compressed video transport requirements

From: andre@rail.mentor (Andre' Hut)
Newsgroups: rec.video.satellite,misc.consumers,misc.consumers.house
Subject: SkyPix cooks up home satellite dish for $700
Message-ID: <ANDRE.91Feb6104313@rail.mentor>
Date: 6 Feb 91 17:43:13 GMT
Sender: Unknown@caeco.UUCP
Organization: /net/rail/home/andre/.organization
Lines: 170


[Reprinted without permission from Electronic Engineering TIMES, Jan. 21, 1991,
by Richard Doherty]

Skypix brings home 80 channels of digitized TV
- ----------------------------------------------
Two-foot dish, set-top digital decoder will be priced at $699

     -24-inch parabolic reflector
     -Cassegrain signal mirror
     -Ten channels of digitally compressed programming
     -Ku-band GaAs downconverter (12Ghz)
     -Infrared Remote
     -Phone connection for credit ordering
     -Stereo, 480-line NTSC video

Las Vegas, Nev. -- The cable TV and backyard satellite dish industries are
girding for a new challenge: the first window-mount home satellite system
available in the U.S.

SkyPix (Kent, Wash.) offered a sneak preview of its compact dish system at
last week's Consumer Electronics Show here and will be on hand at this week's
Satellite Business Communications Association show, also to be held here.  The
startup expects to begin selling systems through retail channels this summer
for about $700.

SkyPix hopes to tap digitial compression technology licensed from Compression
Labs Inc. and advanced Ku-band digital RF modulation to deliver home
programming at a price that will challenge existing cable TV and larger C-band
satellite systems.

Initially, SkyPix will offer up to 80 channels of video programming, but the
system's decoder, a set-top box based on three VLSI chips, will be able to
receive up to 250 channels when more powerful satellites are launched later
this decade.

...

The SkyPix system achieves a data-compression rate of 54 to 1 by using Sun
Microsystems Inc. computers operating at 150 Gflops along with digital-
compression technology licensed from Compression Labs Inc.  With compression,
each channel requires just 2.2 Mbits/second for video and audio, down from a
source data rate of 120 Mbits/s.  Error-correction overhead brings the total
channel requirement to 3 Mbits/s.

The 10 Ku-band channels that are decoded from an RF tuner section that includes
three custom LSI chips.  These chips that the 160 Mhz fo 80 channel data and
extract specific TV images, stereo sound and data.

SkyPix said it will purchase Ku-band GaAs down-converter assemblies from a
number of sources.  These devices change the 12-Ghz incoming satellite signal
into a UHF-band signal that is easier to convey on low-cost coaxial cable.

Once the antenna is aimed at the Hughes SBS-6 satellite, it can be mounted
under the eaves of a house, on its roof or side or on the ground.

At 12-Ghz Ku-band frequencies, satellite antennas usually are subject to one
nagging problem: signal absorption during rainfalls.  In the satellite
industry, signal fades tied to local downpours are referred to as rainouts.
But because of advanced digitial modulation and error-correction technology,
rainouts are not expected to be a problem, according to SkyPix.

The SkyPix system sports a signal reserve of 5db.  In severe cases in which
total signal interruptions occur -- such as when an aircraft flies over the
antenna -- the system will lock into a digitial freeze-frame until the signal
is restored.

In demonstrations viewed by EE Times at the Winter Consumer Electronics Show
last week, most of the video channels shown displayed better than TV-broadcast
quality.  Eight channels were shown operating at one time, fed by a 36-inch
satellite dish.  Image color depth was excellent, and video resolution was
above NTSC standards.

However, during some fast-moving action scenes, there were moments when the
digital compression could not keep up with screen refresh demand.  At these
times, much of the picture turned grainier, as the digital compression tried
to keep up with deltas from one frame to the next.  Overall, the images were
free from RF signal ghosting and coaxial reflection that plague some cable
TV systems.

     -- Richard Doherty
- --
- -------------------------------------------------------------------
Andre' Hut       {mntgfx,utah-cs}!caeco!andre
Mentor Graphics, Suite 300, 5295 South 300 West, Murray, Utah 84107
- -------------------------------------------------------------------

------- End of Forwarded Message