[XviD-devel] Quality optimization

[Marco Al]

Michael Militzer wrote:

> yes, vhq mode especially seems to help with 4MV mode. INTER4V mode does not
> give much benefit currently, maybe we can improve this

If you go this way I think a fine-to-coarse search would give better results
than the present coarse-to-fine search.

> why hadamard and not DCT? (because it's faster?) - shouldn't DCT give better
> results (at least in theory)? And why SAD and not SSE? SSE should more
> evenly distribute the error over a whole block which should make it easier
> to code.

Neither SAD nor SSE evenly distribute the error across the block if you use it
in the transform domain AFAICS (although admittedly in the spatial domain SSE
should do that better than SAD, as you can see in Dirk Farin's paper "A New
Similarity Measure for Sub-Pixel Accurate Motion Analysis in Object-Based
Coding" it doesnt do it all that well either though).

Marco