[XviD-devel] some bframes ideas

[Radek Czyz]

> Sorry, I may be a mathematician, but this is too fast for me.
> Do you want to modify dequantization?

> Yep, as I said ... during dequantization.

Ooops, that's not what I thought ;)
I'll test 'my idea' anyway >:)

So, one more time, what I wanted to do: decrease magnitude of all
coefficients by, for example, 1/4 of maximum quantization error. This
will decrease PSNR, but will also decrease bitcount, so the maximum
quality at fixed bitrate might be somewhere there.

Can someone tell me if I can obtain the 'maximum error' matrix by
creating a matrix of all 1s and dequantize it? Or maybe it will give
me double error (1.5 of an error?) because of this bias towards zero
during quantization?
Sure, I can just analyze quantization code and I'll know ;)

Radek