[XviD-devel] other FDCT with fewer multiplications

[Christoph Lampert <chl@xxxxxxxxxxxxxxxx>]

Hi,

I got an email with a paper and a source archive for a different
method of DCT, including possibility scaling, which works with less
multiplication than usual column/row. I attached the source file,
that currently uses doubles, so it's slow.

Maybe somebody (Skal?) can take a look if this is worth testing to MMX? 
I can send the paper in PDF, too, if it's needed.

Christoph

2dDCT.zip
Description: Zip archive