[XviD-devel] Hadamard Transform
Christoph Lampert
xvid-devel@xvid.org
Thu, 5 Sep 2002 14:03:47 +0200 (CEST)
Hi,
since so many people seem to be good at programming and ASMing (and I am
not): Could you try and tell me how much work calculation of
Hadamard's transform (e.g.
http://www.iro.umontreal.ca/~pigeon/science/ondelettes/Hadamard/Hadamard.html
)
is compared to fDCT and/or SAD?
Hadamard is used in H.26L as apporoximation for DCT, when information on
transformed coefficients is needed, but DCT itself is too slow
(e.g. mode decision). If it's good and fast, it would come handy in
advanced adaptive quantization, too.
gruel