[Mp4-tech] [Video] H.264: What to do in case of a P-frame loss?

[Luqman]

hi Shankar,
Thank you for your valueable posting. I will  tweak configuration files
of encoder and decoder to find the most suitable settings for my
purpose.
Regards,
Luqman 
> Shankar Manuel Aghito wanted us to know:

>Dear Luqman and Sugeeth,
> 
>just to further clarify. The standard  describes how an intact bitstream is decoded.
>A conformant decoder is allowed to crash if there is any error in the bitstream
> 
>Fortunately error robust decoders can be implemented by using error concealment techniques (i.e. filling the holes!!). Some error concealment (probably not the most advanced) is implemented in the reference software (different algorithms are utilized wheter only a part of a picture or the whole picture is lost - i think that this was previously discussed in this mailing list, including the links to the documents describing the algorithm - you should be able to find it). Implementing more advanced error concealment algorithms is a way for companies to differentiate their products.
> 
>As you said, the reference decoder support error concealment in the case of loss of NAL units. Bit error concealment is not implemented, i.e. if one bit is damaged you should remove the whole NALU.
> 
>In the encoder there are several things that can be done do make easier the job of the decoder, for example forcing intra pictures/macroblocks for stopping temporal error propagation, using multiple slices per pictures and eventually flexible macroblock Ordering (FMO) in order to limit the part of the picture that is lost.  These things can be done using the reference encoder. Another possibility is using redundant slices ( I didn't test it personally, but an implementation of redundant slices should be included in the latest version of JM).
> 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 191 bytes
Desc: Digital signature
Url : /pipermail/mp4-tech/attachments/20061018/c10e0186/attachment.bin