[XviD-devel] VHQ

[Radek Czyz]

> I guess it would be logical then not to use "bits", but differences
> between _unquantized_ DCT data, and since DCT in theory is an orthogonal
> operation, that should be equal to SSE with SAD acting as an approximation
> of this.

I also thought about checking total quantization error. In this case,
this would do what we need.
However, my tests with dequant->count error were horribly slow, and in
fact it didn't help a thing.
Also, we would have to create some relations between total error and
bits (how much of an error should equal one bit?) and the whole search
would not be so pure, nice and clean.
I think I like the way it is now - if cbp = 0, we count on SAD. If
not, we count bits.

Radek