[XviD-devel] dark blocks in b-frames

[Radek Czyz]

lo,

> now at dequantization, we multiply all the non-zero coefficients by 14
> _and_ the add 13 (quant-1). the dequantized matrix then becomes:

You substract 13 here ;). I really hope you mean "+/- (quant-1), away from
zero"

>          -27, 0,  0,  0,  0,  0,  0,  0,
>           0,  0,  0,  0,  0,  0,  0,  0,
>           0,  0,  0,  0,  0,  0,  0,  0,
>           0,  0,  0,  0,  0,  0,  0,  0,
>           0,  0,  0,  0,  0,  0,  0,  0,
>           0,  0,  0,  0,  0,  0,  0,  0,
>           0,  0,  0,  0,  0,  0,  0,  0,
>           0,  0,  0,  0,  0,  0,  0,  0,

Ok, so quantization rounds towards zero, and then dequantization
assumes that the rounding was as bad as possible? That's weird. I
don't get it. Why 13, when the average error is 6? MPEG4's weird.

Anyways, in light of my previous mail - I think we'll have to detect
this kind of blocks and 'fix' them, simply because they look horrible.

> as gruel has said, this probably isnt a bug, rather the effect of using
> a mix or high quantizers for bframes.

I wonder why noone noticed this in DivX5. They should also have the
problem, especially because they use even higher quantizers for
bframes.

Radek