[XviD-devel] Some mismatch in chroma

Michael Militzer michael at xvid.org
Wed Nov 29 10:01:15 CET 2006

Hi Yousuf,

we don't use Vprove but instead test against the MoMuSys and MSFDAM MPEG-4
reference implementations. That said, you cannot expect bit-identical
output from two compliant decoders for the same input clip. That is because
e.g. for idct two compliant decoders can use a different implementation
with quite different error specs - the standard is not very strict here and
mainly just requests a IEEE1180-compliant idct implementation.

In result, mismatch can occur that can accumulate and exhibit noticeable
artifacts under some circumstances even with compliant decoders. So the +-5
diff your report sounds quite large but could be possible - did you observe
that the diff becomes larger and larger in-between keyframes or is it
constant? Do you see such large mismatch for I-frame only video too?

You can certainly send me your stream and yuv output too (if it's not too
large) if you like...


Quoting Yousuf Ali <y0u5uf at yahoo.com>:

> Hi guys,
>   New member here, first email so please forgive me if I am not   
> doing this properly.
>   To get straight to the point, I have a compressed (raw, m4v) mpeg4  
>  video stream that I decoded with the xvid 1.1.2 codec (using   
> decraw). I also have an mpeg-4 analyzer (at my workplace) called   
> Vprove (from Vqual) and I decoded the same stream with that.
>   I see some mismatch in the chroma channels. The errors are small   
> though, around +/- 5 in U and V channel outputs as far as I saw.
>   I want to know if this is a known fact? If it is, then does this   
> mean that xvid is not 100% compliant? It could even be the case that  
>  Vprove might be wrong, question is, how do we know?
>   If you guys are interested to analyse this further, I can send you  
>  the compressed stream, source YUVs and the analyzer trace and   
> decoded YUV (from Vprove) for this particular stream.
>   On a side note, can I ask you what is the method of testing you   
> employ to verify compliancy? Since the errors are small enough, not   
> really noticable in output picture, so does that "good enough" mean   
> compliant enough as well?
>   I only ask about the testing methodology because I am working on   
> an mpeg-4 encoder model myself (at home) and up until now, my main   
> validation tool was xvid (decraw). And now I see a stream generated   
> by my encoder that matches Vprove but not xvid :(
>   Regards,
>   Yousuf
> ---------------------------------
> Access over 1 million songs - Yahoo! Music Unlimited.
> _______________________________________________
> XviD-devel mailing list
> XviD-devel at list.xvid.org
> http://list.xvid.org/mailman/listinfo/xvid-devel

More information about the XviD-devel mailing list