Log in

View Full Version : Technical question about quantizer


hellgauss
31st January 2005, 12:23
What is 'quantizer'? And how it works in the encoding process?

I've read in some guides that the encoding process is done in this way:

1)The codec takes an 8x8 block
2)It does the dct and trasform it in a 8x8 frequency block
3)Each coefficent is divided by the coefficent of the quantization matrix (h263, mpeg, ecc....)
4)The result is stored in a compressed stream (like a zip file).

Where does the quantizer does his work? What is exactly the differences between a q3 frame and a q4 frame?

Thanks and sorry for bad English

HG

[EDIT]
sorry... i've spent three days searching an answer in this forum, and i didn't see that just yesterday someone asked the same question....

Koepi
31st January 2005, 12:56
That's quite simple to answer.

The quantisation matrix values get multiplied by the quant scale factor (quantiser). So more more pixels of the 8x8-block get zero'ed (=loosing details). Motion vector length gets also limited with rising quantiser.

I hope this helps.

Cheers
Koepi

sysKin
31st January 2005, 13:07
Originally posted by Koepi
Motion vector length gets also limited with rising quantiser. Nope ;)

Koepi
31st January 2005, 14:02
Huh? I thought fcode = MV length. My bad, sorry.

Cheers
Koepi

sysKin
31st January 2005, 15:50
Originally posted by Koepi
Huh? I thought fcode = MV length. My bad, sorry.
Oh yes, fcode = MV length but quantizer != fcode ;)