[XviD-devel] VHQ

[Marco Al]

From: "Christoph Lampert" <chl at math.uni-bonn.de>

> So what we would need is a criterion _when_ the bits saved actually
> outweight the residual error, and it seems "needed bits" alone is not
> enough for this :-(  Current implementation of search uses some kind of
> langrangian with fixed lambda-value (depending on quantizer), but this
> doesn't seem to be perfect, either, of course.
>
> Now what?

I havent taken a look at ffmpeg's code, but it says its dct based ME is R-D
optimized ... maybe take a look?

A lagrangian multiplier technique where you update the multiplier based on
history and the rate targets from rate control (not easy) might work, and
shouldnt be noticably slower than the pure rate based search (apart from the
fact that there is no early out from the SAD based search).

Marco