[FFmpeg-devel] dealing with tables in DV codec

Roman V. Shaposhnik rvs
Thu Sep 11 04:58:20 CEST 2008

On Thu, 2008-09-11 at 04:51 +0200, Stefan de Konink wrote:
> If the motion estimator tells the encoder that some macro/superblocks
> are equal between frames, would you still have to redo the complete
> procces because of the shuffeling and dct?

Well, the comparison was strictly of a decoding speed. But to answer
your question -- DV bit-encodes an entire super block (5 macro blocks
taken from very different parts of the frame) thus without trickery
it seems that an entire superblock has to remain motion-less for this
optimization technique to work. Besides, I'm really curious as to
how much slower the [i]DCT is vs. motion estimation (or even bit


More information about the ffmpeg-devel mailing list