Re[2]: [XviD-devel] xvid + bframes and mplayer/ffmpeg decoding

[Marco Al]

Christoph Lampert wrote:

> I explained it two days ago in another mail. For a different decoder
> the result might be non-zero, and when such a block is decoded several
> times in a raw, the difference between the encoders internal image and
> the decoders image grows.

Yeah but you didnt quite describe why it would be more likely to let
iDCT errors acccumulate :/ Took me a while to realise that while with
motion the error would be distributed to other frequencies (unless
the motion is exactly a multiple of 8 pixels) and end up as noise just
like all the other iDCT errors which are allowable ... but that for MBs
with completely static content it was quite likely to happen.

Marco