[Mp4-tech] Practical Issues In H.264

[Dzung Hoang]

I suggest reading the following research paper from IBM.
http://www.research.ibm.com/journal/rd/434/gonzales.html
Regards,
- Dzung Hoang
________________________________________
From: mp4-tech-bounces lists.mpegif.org
[mailto:mp4-tech-bounces lists.mpegif.org] On Behalf Of Satendra
Sent: Tuesday, September 26, 2023 2:12 AM
To: Gary Sullivan
Cc: Sugeeth; mp4-tech lists.mpegif.org
Subject: Re: [Mp4-tech] Practical Issues In H.264
Hi Gary,
when we say that the quality of encoded bitstream and hence the quality of
the decoded Video sequence, is in some way depends on the search range and
the way we are implementing it, like where it is centred. 
I want to know what can be  more criteriae to select the search ranges, I
can count few... please point out if I am wrong,
1) Image size
2) Quality
3) Complexity of target encoder/decoder.
Actually I want to know what are the optimal search ranges, and what are the
most common settings generally found in encoders in this regard. Will the
neighbouring blocks and MBs will have MVs pointed to this search range only?
Thanks
Satendra
On 9/25/06, Gary Sullivan <garysull windows.microsoft.com> wrote: 
Sugeeth et al,
When considering the demands of a simple input video sequence versus a 
more difficult one, you need to consider both the picture quality and
the bit rate.  The more difficult the scene, the more bits will be
needed to achieve a given level of quality.
I would guess that most applications probably operate somewhere in the 
range of 34 to 41 dB PSNR.  That covers a significant quality range.
Different applications have different levels of quality that are
considered acceptable.
Different people have different interpretations of the concept of a 
motion search range.  In some cases the range refers to a searched area
around the (0,0) vector value.  In other cases it refers to a different
searched area.  The second of these two characterizations is the
relevant one in that reference software, and this is may be why you do
not find larger search ranges to be making a big difference in quality.
You did not say whether you are running the software with the "rate 
control" feature enabled or not.  If you are not using rate control,
then the quality of the pictures decoded from the bitstream will
probably be approximately the same (in PSNR terms) regardless of the
input video data characteristics or the motion search range.  What will
vary is the bit rate, not the quality, unless you enable rate control or
vary the value of the quantization parameter.
In regard to your question about what tools are available "in h264" to 
measure quality -- this is not a proper question, since the standard is
concerned only with video representation - not quality measurement.
Perhaps you're referring to the software rather than the video coding
spec itself.  In that case I think PSNR is the only quality measure that
is built into the software.
Keep studying.
Best Regards,
Gary Sullivan
+> -----Original Message-----
+> From: mp4-tech-bounces lists.mpegif.org
+> [mailto:mp4-tech-bounces lists.mpegif.org] On Behalf Of Sugeeth
+> Sent: Monday, September 25, 2023 3:15 AM 
+> To: mp4-tech lists.mpegif.org
+> Subject: [Mp4-tech] Practical Issues In H.264
+>
+> The ISO-Reference Code when run for a vector like stefan
+> doesnt show too 
+> much quality in the reconstructed frame ( And also the h.264
+> passed thru
+> the decoder) when compared to a simple vector like akiyo or claire.
+> Stefan is basically complex because, there is a huge amount 
+> of movement
+> and uncertainity. Well, In real time, a moving camera would
+> also face
+> certain complex issues if it even has a h.264 codec in it. So, what
+> could be the max. possible psnr in such a complex case 
+> considering that
+> all the tools of the h.264 standard are kept as best as
+> possible????????
+>
+>
+> P.S: I observe that in Stefan , since there is a lot of
+> sudden movement 
+> involved, i presume , that the parameter affecting the
+> vector,  maximum
+> would be search range,  however, there is not much of a quality
+> difference beyond a search range of 16. i.e a search range 
+> of 16 or a
+> search range of 128 give me just about the same quality. and other
+> parameters like bitrate , intermode  have been allocated
+> sufficently?????
+>
+> In general, what is an acceptable value for psnr in h.264???? Is it
+> vector dependent or is there some value  of psnr that determines the
+> quality in h.264???? Other than Psnr, what are the other
+> tools available
+> in h264 to measure quality???? 
+>
+> thanks in advance,
+> sugeeth .
+> _______________________________________________
+> NOTE: Please use clear subject lines for your posts. Include
+> [audio, [video], [systems], [general] or another 
+> apppropriate identifier to indicate the type of question you have.
+>
+> Note: Conduct on the mailing list is subject to the
+> Antitrust guidelines found at
+> http://www.mpegif.org/public/documents/vault/mp-out-30042-Ant
+> itrust.php
+>
_______________________________________________
NOTE: Please use clear subject lines for your posts. Include [audio,
[video], [systems], [general] or another apppropriate identifier to indicate
the type of question you have. 
Note: Conduct on the mailing list is subject to the Antitrust guidelines
found at
http://www.mpegif.org/public/documents/vault/mp-out-30042-Antitrust.php 
-- 
----------------------------------------------------------------------------
------------------------------------------------------- 
"We all agree on the necessity of compromise. We just can't agree on when
it's necessary to compromise."  ------Larry Wall 
----------------------------------------------------------------------------
-------------------------------------------------------