[opendtv] Re: AVC realistic bitrates?
- From: Craig Birkmaier <craig@xxxxxxxxx>
- To: opendtv@xxxxxxxxxxxxx
- Date: Thu, 9 Jun 2005 07:55:14 -0400
At 1:21 PM -0400 6/8/05, Manfredi, Albert E wrote:
> What makes AVC more efficient, in large
>measure, is that interframes don't need to be transmitted
>as often. But not transmitting interframes as often has
>an impact on how fast one can lock onto a stream. So it's
>very important to assign the GOP as required first,
>before comparing quality vs bit rate.
Your really reaching on this one Bert.
Longer GOPS help both MPEG-2 and AVC in terms of efficiency. This has
been documented many times. There are at least a dozen reasons why
AVC is more efficient, and I frame frequency is not one of them.
Furthermore, the need for I frames twice per second is questionable.
Clearly if your main concern is channel acquisition time, then
shorter GOPS are desirable. But GOP length is not critical for
downloaded files - you can use whatever works best for the source
material, just as is done today by compressionists when creating DVDs.
>That would be a mistake. My first question is why don't
>you compare MPEG-2 at 1 or 2 Mb/s compared with AVC at
>those rates? And my second question would be why you
>expect to derive anything of value about quality at the
>higher bit rates from this? After all, just because AVC
>at 1 Mb/s looks okay, this proves nothing about AVC vs
>MPEG-2 at higher bit rates.
And the answer is that MPEG-2 at 1-2 Mbps is only possible if you do
a lot of image processing first. When you do this you will end up
with soft - often cartoon like - images, or so many compression
artifacts as to render the content unwatchable.
The advantage of AVC have been documented at a wide range of bit
rates. Gains in efficiency have been noted across all bit rates and
types of material. The relative gains vary, but the average is at
least 2:1 and sometimes higher. And as others have noted, we are just
beginning to learn how to use the AVC tools, while we have already
maxed out what is possible with MPEG-2.
>You must be joking. There's no way the 1 Mb/s examples
>qualify as HD. Maybe we should compare with SD MPEG-2
>images. Without further info, I can believe that AVC at
>1 Mb/s compares favorably with MPEG-2 at 2 Mb/s, maybe
>even 3 Mb/s. AVC is supposed to hold up very well at the
>low bit rates, unlike MPEG-2 which falls apart below
>~1.5 Mb/s.
When did HD enter this discussion. Clearly it is inappropriate to
compare apples and oranges.
AVC has significant advantages when comparing SD to SD and HD to HD.
And Bert is correct that it does a much better job when the bit rate
is severely constrained. But it also does a much better job when the
bit rate is unconstrained, as is the case for compression for
acquisition and digital cinema applications.
I saw a demonstration of HD source compressed using AVC with the
Fidelity Range extensions at NAB that was very interesting. At bit
rates above 100Mbps I could not see any degradation of the HD source
(note that the bit rate for Sony's HDCam is above 150 Mbps and DVCPro
HD is around 100 Mbps). At 50 Mbps I was just beginning to see some
artifacts.
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:
