[opendtv] Re: MPEG-2 et al
- From: "Tom McMahon" <TLM@xxxxxxxxxx>
- To: <opendtv@xxxxxxxxxxxxx>
- Date: Mon, 24 Jan 2005 01:55:44 -0800
One thing I've come to appreciate over the years is that the output quality of
a video codec standards body is only as good as the
test content that it uses to develop the respective standard, and only then
across the range of resolutions and bitrates that were
used in the development process.
In any of these activities, there are numerous proposals on how to tweak and
tune the "design" for greater efficiency and/or
quality. Proposals are accepted into the standard or rejected based on the
numbers and an ability to have some else reproduce your
results. However, sometimes the proposed algorithms (or "tools") are highly
content dependent (or only work at higher bitrate, for
example). That which shows great results for interlace may not work well for
film grain. And so on.
In the development of MPEG-2 people were groping in the dark in this respect.
When MPEG-2 was developed there was no digital cable,
there was no digital satellite, there were no DVDs, there was no HD. The
content used to develop and tweak the Standard was ad hoc
at best.
H.264/AVC was designed to serve many masters and the test cases were explicitly
manifold (in contrast to the above). I think we did
a pretty good job of bashing the mathematical concepts against a severe range
of representative market-based content during the
development process. That is why we're seeing 1) such good coding results and
2) why H.264/AVC will only get that much better as
ENcoder manufacturers better-employ the range of tools we've built into the
standard.
-----Original Message-----
From: opendtv-bounce@xxxxxxxxxxxxx [mailto:opendtv-bounce@xxxxxxxxxxxxx] On
Behalf Of Tom Barry
Sent: Sunday, January 23, 2005 4:17 PM
Manfredi, Albert E wrote:
> Can't you do that already with MPEG-2? For > example, isn't that exactly
> what happens when > MPEG-2 is used in a fixed bit rate
mode? You > get variable quality.
I think you get variable quality but only by decreasing the precision of each
frequency component of a block, not by decreasing the
resolution and total number of blocks sent. For instance, if you send 720p in
MPEG2 at only 5-6 megabits it will likely look worse than if you instead scaled
it down to 480 lines (or less) and scaled it back up
for display.
It seems there is sort of a sweet spot resolution for each source and bit
rate. So it is probably better to vary both the
quantization and resolution IMHO.
- Tom
> Tom Barry wrote:
>
>
>>Thus I'd predict we eventually will have
>>compressed video that has a MAX resolution
>>but not a fixed one and we can then just
>>rate shape everything from a single source
>>to multiple targets such as sub-channels
>>and cell phones. Each step along the way
>>might discard a bit more as needed. But
>>the master archives would be pre-encoded at
>>highest quality
>
>
> Can't you do that already with MPEG-2? For
> example, isn't that exactly what happens when
> MPEG-2 is used in a fixed bit rate mode? You
> get variable quality.
>
> Bert
>
>
> ----------------------------------------------------------------------
> 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.
>
>
----------------------------------------------------------------------
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.
----------------------------------------------------------------------
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:
