[Mp4-tech] FW: MPEG-4 part 2 reference software bug: encoder and decoder outputdifferent motion vectors

[MPEGIF List Admins]

This was sent to mp4-tech-bounces instead to mp4-tech. Forwarding.
  _____  
From: Tung yi-Shin [mailto:tung cmlab.csie.ntu.edu.tw] 
Sent: Sunday, 22 October 2023 06:08
To: mp4-tech-bounces lists.mpegif.org
Subject: RE: MPEG-4 part 2 reference software bug: encoder and decoder
outputdifferent motion vectors
Hi tomer, 
Could you try the latest version of MoMoSys code instead? The latest version
is FGS-Ref-Software-FDAM-N4711-MoMuSys.zip (FDAM), which is provided in
14496-5:2001/AMD 1. Many fixes are included in this version. If the problem
is still there, please send your encoding parameter files to me? I would
trace it and provide fixes for you if it is indeed a bug.
Regards,
Yi-Shin
From: mp4-tech-bounces lists.mpegif.org
[mailto:mp4-tech-bounces lists.mpegif.org] On Behalf Of Tomer Segal 
> Sent: Tuesday, August 01, 2023 7:33 PM 
> To: mp4-tech lists.mpegif.org 
> Subject: [Mp4-tech] encoder and decoder output different motion vectors 
> 
> 
> hi, 
> i'm a little confused about a set of results i got using the mpeg4
reference software (MoMuSys-1.0-001220_sony). 
> i encoded a yuv sequence having the encoder output the motion vectors
calculated during encoding. then decoded the resulting mpeg file with the
decoder, enabling a trace of the motion vectors extracted during  decoding.
i repeated these steps twice, once for a very simple yuv sequence (few
blocks change throughout the sequence) and a second time using a more
complex sequence. for the simple sequence the motion vectors output by the
encoder and the decoder were exactly the same. for the complex sequence
however there were differences between the vectors - in most blocks vectors
differed by a single halfpel, but a few of them differed by as much as 5
halfpels. 
> i thought the motion vectors extracted by the decoder were suppose to be
exactly the same as the ones calculated in the encoder. am i wrong? i tried
finding the answer in the mpeg4 iso and other related articles but i was
unable to. 
> if indeed the vectors are supposed to be identical, what could be causing
this difference? 
>   
>   
> Thanks, 
>     Tomer. 
>   
>   

-- 
Communication and Multimedia Laboratory 
Dept. of Computer Science and Information Engineering, NTU 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: /pipermail/mp4-tech/attachments/20061022/4df153f4/attachment.html