Quantization table...

[octapus]

In the new Xvid 2-3-2002 binaries there is somewhere an always disabled button with "Quantization table" Text on It. Is this going to be the next immediate feature of xvid?

[Acaila]

Hey -h, here's your first! :D

That option was disabled on purpose, because it doesn't have a function for us.

[Klumsy]

Here is what Koepi's guide says about it:

"The Quantization Matrix... button is for future use, therefore it's disabled now. This option could render
the modulated quantization type useless, because you can build your quantization matrix accordingly
to that :)"

Don't know exactly what it means but it sounds good :D

[ProfDrMorph]

as you all know XviD offers different quantization types ( MPEG, H.263 and Koepi's build also offers Modulated quantization ). I don't know the exact differences between them but what they all do is using a quantization matrix to quantize the output of the DCT algorithm. I don't know about MPEG but old image compression algorithms just divide the DCT output matrix by the quantization matrix.

BTW: TMPGEnc offers you to use your own quantization matrix.......

[Koepi]

As I posted on another thread (right below this one ATM), tomorrow suxxen_drol will commit it to CVS. We then may incorporate it into the GUI (well, just activate it and use the API to send it to the core).

This will break "my" patches, so you'll have to decide whether you want modulated_quantization_type or your own matrix...

Regards,
Koepi

[ProfDrMorph]

why does it break your patches? What "Modulated" does is, deciding whether to use MPEG or H.263 on a per-frame basis, doesn't it? So maybe you could let the user decide which quantization matrix/type to use for low quantizers, a break point ( let's say quantizer = 4 ) and another quantization matrix/type for quantizer that are equal to or bigger as the break point. Am I getting something wrong here?

[octapus]

I found hardcoded inside the source code this:

static const int16_t default_intra_matrix[64] = {
8,17,18,19,21,23,25,27,
17,18,19,21,23,25,27,28,
20,21,22,23,24,26,28,30,
21,22,23,24,26,28,30,32,
22,23,24,26,28,30,32,35,
23,24,26,28,30,32,35,38,
25,26,28,30,32,35,38,41,
27,28,30,32,35,38,41,45
};

So is this the standard quantization table that xvid uses for intra frame quantization?

I'm asking because i saw something strange (concerning this) in the latest build:

in the new code that has been inserted for quantization table there is this declaration

BYTE default_qmatrix_intra[] = {
8, 16, 19, 22, 26, 27, 29, 34,
16, 16, 22, 24, 27, 29, 34, 37,
19, 22, 26, 27, 29, 34, 34, 38,
22, 22, 26, 27, 29, 34, 37, 40,
22, 26, 27, 29, 32, 35, 40, 48,
26, 27, 29, 32, 35, 40, 48, 58,
26, 27, 29, 34, 38, 46, 56, 69,
27, 29, 35, 38, 46, 56, 69, 83};

And the default intraframe quant. table is different and it's the same with default of TMPEGEnc for intra.

[-h]

I didn't think filling the "custom" matrix with the standard values made much sense, so I just stuck the TMPGEnc matrix in as the default for a bit of variety.

You'll probably see good results from bringing the MPEG4 standard matrix closer to 32 for all the values - this will be a median between H.263 and MPEG quantization.

-h