[XviD-devel] lumimasking problem

[Radoslaw 'sysKin' Czyz]

Hi everyone.

I was investigating bugs with adaptive quantization and my hinted ME
search.
I was dumping quantizer data at every interaction of motion estimation
and I discovered ome very strange things..... Anyway, I have to assume
that my ME doesn't work with lumimasking at all, because quant data
are different then I had expected:
Strange things:
    - first several frames don't have adaptive quantization at all. It
is strange but at least doesn' hurt.
    - mostly quant/dquant are ok
    - now, look a at his debug info. It's all one frame:

dquant = 64, quant = 3
dquant = 64, quant = 3
dquant = 64, quant = 5
dquant = 3, quant = 6
dquant = 2, quant = 6
dquant = 64, quant = 6
dquant = 64, quant = 6
dquant = 64, quant = 5
dquant = 64, quant = 3
dquant = 64, quant = 3

where NO_CHANGE is 64. Weird, isn't it?

or this:
dquant = 64, quant = 4
dquant = 64, quant = 4
dquant = 64, quant = 4
dquant = 3, quant = 6
dquant = 2, quant = 8
dquant = 64, quant = 8
dquant = 64, quant = 8

I don't know if "2" or "3" means +2 but only one can, right?

I'd be grateful if someone who knows more about it could comment this.

Radek