[Mp4-tech] RE: H.264 Buffer analysis

[yashwinder chhikara]

Hi Gary,
You are absolutely right in saying that the VUI and timing information should 
be available to test the conformance, either they may be presesnt in the stream
or that information should be available by some other means which are not
defined.
So i wanted to know is there any work going on in this direction of defining these
other means from where we can get this information and how?
If that information is not available can we check the conformance using any 
algo. which is defined?
Could you give any pointers to the tool youmention which,  has the capablity of scanning a stream and computing the convex hull of the curve relating the necessary buffer capacity and bit rate for decoding the stream.
Thanks 
Yashwinder
Gary Sullivan <garysull windows.microsoft.com> wrote:
There are several issues here.  There is no requirement that the VUI and timing information be in the stream.  There is, however, a requirement that such information be available in order to test the conformance of a stream.  There is a difference between having the information be "present" and having it "in the stream".  Any way that the information can be provided is fine as far as the standard is concerned.  One such way, for example, is for the timing to be known to be locked to a particular system frame rate.  The timing information can be provided by the system rather than within the video syntax.  Many system designs would carry such information -- video is almost never used all by itself without audio and a system to synchronize everything.
Streams without that information provided one way or another could be useful for testing output-order conformance to the standard for decoders, but not for testing output timing conformance to the standard.
We have also determined that it is ordinarily conceptually possible to construct timing information that can be associated with any stream that has that information missing, and to do so in a way that makes the stream conform, unless the stream contains obvious fundamental problems such as a hugely excessive number of bits used for an individual picture.  Some time ago, we had a tool that was part of the reference software package that was capable of scanning a stream and computing the convex hull of the curve relating the necessary buffer capacity and bit rate for decoding the stream.  I think that tool became obsolete as minor details were later changed, but the concept remains valid.  It would be nice if we could resurrect that kind of capability.
Probably all or nearly all of the streams that have been provided without that information are actually designed to be conforming with a typical fixed frame rate such as 30 or 15 or 10 or 25 frames per second.
Ideally, we should really have that data for all of the streams.
Best Regards,
Gary Sullivan
---------------------------------
From: yashwinder chhikara [mailto:y_chhikara yahoo.com]
Sent: Tue 5/25/2004 11:35 PM
To: Sriram Sethuraman; Gary Sullivan; Iain Richardson; jvt-experts mail.imtc.org
Cc: mp4-tech lists.mpegif.org
Subject: H.264 Buffer analysis
Hi,
 In H.264, for doing the CPB buffer analysis, We require the VUI information, the buffering period and picture timing SEI messages to be present in the stream. If this information is present then we can do the required CPB buffer analysis as given in the specs.. Now if this information is not present in the stream ( which is the case for approx. 95% of the test streams available), can there be any other manner we can do the CPB buffer analysis using the coded information. 
Any ideas are welcome... 
Regards 
Yashwinder 
---------------------------------
Do you Yahoo!?
Friends.  Fun. Try the all-new Yahoo! Messenger
-------------- next part --------------
An HTML attachment was scrubbed...
URL: /pipermail/mp4-tech/attachments/20040526/79dc3b52/attachment-0001.html