[XviD-devel] Re: vfw 2pass stats

[Marco Al]

Marco Al wrote:
> Marc FD wrote:
>
>> Maybe i'm wrong, but wouldn't be better to make ONLY a vbr mode in the core.
>> Not a scaler, only the stuff we need to encode vbr like (x-pass).
>
> You can do scaling in vfw. If you wanted to do something a little more
> sophisticated like using the R-D curve, which is only known to core in the
> second pass because of the reasons Michael mentioned, to determine the number
> of bits for a given frame you will have to do it in core though.

I just noticed Zhihai He et al just recently published a paper on 2 pass
encoding using the per frame R-D curves from the previous pass (this could be
done outside core). Considering the almost complete lack of research on 2 pass
encoding this should be interesting to a couple of people here.

The paper is titled "Optimal bit allocation for low bit rate video streaming
applications". It is in the proceedings of ICIP-2002.

Marco