[opendtv] Re: PR: Analog Devices' JPEG2000 IC Enables Wireless High-Definition Video Distribution in the Home

At 1:42 PM -0400 8/24/05, Manfredi, Albert E wrote:
>What they say is true for M-JPEG as well. Big deal.

The big deal is that interframe compression IS NOT the best choice 
for all applications. It adds computational complexity, latency, and 
cost.  If there is sufficient network and/or storage bandwidth, then 
an intraframe compression scheme offers significant benefits to many 
applications.

>I'd like to see a compression efficiency comparison between
>this, call it, M-JPEG2000 and MPEG-2 or H.264. And, of course,
>as soon as a JPEG2000-based scheme is modified for efficient
>moving image compression, just as JPEG was modified into
>MPEG-2 last time around,

This would be an apples and oranges comparison and prove nothing. ANY 
Interframe compression scheme would offer higher compression ratios.

JPEG WAS NOT modified into MPEG-2. In fact, MPEG-1 was developed in 
parallel with JPEG, using many of the same core technologies for the 
I frame compression techniques. MPEG-2 was developed AFTER JPEG, and 
most of the new tools incorporated into that standard are related to 
interframe compression techniques and the coding of interlaced images.

The JPEG camp was not looking to get into the motion imaging 
compression business. It was innovative designers at companies like 
Avid, Truevision and Media 100,  who saw the potential of using JPEG 
for intraframe compression of video. They needed a frame based 
compression scheme with low computational complexity and latency to 
create the foundation for non-linear editing system. It just turned 
out that the inexpensive chips being designed for JPEG still image 
compression were able to provide the performance necessary to handle 
30 frame per second. At the time an MPEG-1 encoder required a 
significant portion of a rack of electronics to provide a less 
desirable result. It is important to note that the non-linear 
application requires high quality. It took several years to tune 
M-JPEG to provide real transparency in image quality; the major 
reason for this was the performance of hard disk storage systems (at 
that time), which needed to handle 200-300 kBytes per frame of video 
stored to disk. The first systems from Avid and Media 100 used 1 and 
2 GB disk drives and could only handle about 100 - 150 kBytes per 
sframe to disk.

A comparison that would be meaningful would be to compare the image 
quality and bit rate requirements for M-JPEG versus JPEG-2000. For 
in-home networking and video storage, a system that can deliver high 
quality SD images at about 1-2 MBps would be commercially viable. I 
suspect that JPEG-2000 can do this

>or the evolutionary H.264, that quote
>will no longer apply.
>
>Here's a good introduction into the topic. Start with the
>history, then just work through the subsequent sections.
>
>http://www.amara.com/IEEEwave/IW_history.html
>
>This seems like a truly different approach from MPEG/JPEG, as
>opposed to the much more evolutionary nature of H.264. To
>me, discrete Fourier, discrete Cosine, and discrete integer
>transforms are just variations of the same theme.

Obviously the wavelet transform is significantly different. And yes 
the MPEG standards are evolutionary, as they are riding the leading 
edge of the Moore's Law curve. New tools can be added periodically to 
improve the motion compensated prediction routines, and with H.264 
they moved to an integer transform, which improves accuracy without a 
big hit to performance.

But wavelets have had a difficult time historically competing in the 
Interframe compression arena, because of the need to introduce some 
kind of block based coding to handle the interframe predictions. I 
understand that there has been some progress in this area, but for 
now, wavelets are most promising for intraframe compression 
applications.

>
>>  For those of us with experience in these matters, you
>>  could see another transformation in the works. With the
>>  commercialization of JPEG-2000 chips, it would only be a
>  > matter of time before wavelets started to invade the turf
>>  of MPEG and the world of non-linear editing.
>
>Those of us with experience know that the IP holders of
>H.264 will first want to get their investment back, before
>relinquishing this potentially lucrative market to
>newcomers, eh? But a compression effiency comparison would
>be nice.

Too bad for those IP holders. They cannot force people to use their 
inventions, especially if there is a better, cheaper solution 
available for some applications.

Regards
Craig
 
 
----------------------------------------------------------------------
You can UNSUBSCRIBE from the OpenDTV list in two ways:

- Using the UNSUBSCRIBE command in your user configuration settings at 
FreeLists.org 

- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word 
unsubscribe in the subject line.

Other related posts: