VAQ and MPEG quantization matrix?

KikeG · 30th December 2008, 20:07

I've read that VAQ should only be used with H.263 quantization.

However I've tried using it with MPEG quantization and there's also a size advantage. On a 17 sec. clip of already encoded video of Monsters Inc., encoded at constant q=3, there's a 7'4% size advantage using H.2.63 and a 6,9% size advantage using a MPEG quantization matrix.

As I see VAQ increases quantizers in areas with higher detail. I don't understand why this works any different with a MPEG cqm. Maybe it is because with VAQ quality can suffer more in case of MPEG cqm?

Thanks.

Dark Shikari · 30th December 2008, 20:07

Quote:

Originally Posted by KikeG

I've read that VAQ should only be used with H.263 quantization.

You read wrong.

KikeG · 30th December 2008, 20:23

I don't think I read wrong, however I didn't read it from you. Anyway, thanks for the clarification, it's what I wanted to know.

Dark Shikari · 30th December 2008, 20:31

Quote:

Originally Posted by KikeG

I don't think I read wrong

One could define "reading something that is wrong" as reading wrong...

Sharktooth · 31st December 2008, 04:39

using a custom quantization matrix with VAQ may lead to unwanted results. but that depends on the matrix.

flebber · 31st December 2008, 11:53

Quote:

Originally Posted by Sharktooth

using a custom quantization matrix with VAQ may lead to unwanted results. but that depends on the matrix.

Other than trial and error, is there anyway to predict when a matrix and VAQ combo will cause error?

KikeG · 31st December 2008, 12:12

In case of MPEG quantization matrix, what unwanted results?

Gromozeka · 31st December 2008, 14:39

Mpeg matrix with VAQ give good result, but VAQ with ZSM matrix jawor-a - very bad result
mpeg matrix q3 with VAQ very good too

henryho_hk · 16th February 2009, 11:48

Sharktooth, would you please share your opinion or experience of your MPEG-4 ASP Custom Matrices with VAQ? H.263 is really a bit too soft.

Sharktooth · 16th February 2009, 14:48

@henryho_hk: i no longer use xvid for my encodings since years.

@KikeG: i said CUSTOM quantization matrices (MPEG matrix is not "custom"...).

BigDid · 16th February 2009, 21:52

Quote:

Originally Posted by henryho_hk

Sharktooth, would you please share your opinion or experience of your MPEG-4 ASP Custom Matrices with VAQ? H.263 is really a bit too soft.

Quote:

Originally Posted by Sharktooth

@henryho_hk: i no longer use xvid for my encodings since years...

Hi,

I am actually testing some CQM (custom matrixes) to replace standard matrixes for xvid encoding; more in the different AutoGk threads..

When I will have finished my tests I could share my experience; Atm it is just an opinion

To the point, usually MPEG is a little more compressible than H263 and sharper but H263 is still excellent for anim or 3D-anim
So if you want sharp or more sharp, less compressible than H263 try Eqm V2 (AutoGk sharp) and Eqm V3 HR
If you want more compressible and still sharper than H263 try Eqm V3 LR and/or Eqm V3 ULR

My first tests shows that all CQM from Sharktooth behaves well with VAQ and that gives me the opportunity to thank him for all the worK done

Did

Sharktooth · 17th February 2009, 03:48

that was... unexpected, however im happy they dont screw up...
on a sidenote eqm v3 ULR is obsolete since LCD displays. the reason is the "artificial" sharpening becomes an issue with high-contrast displays.
it is still valid if you have a CRT monitor/TV though...

BigDid · 17th February 2009, 04:05

Quote:

Originally Posted by Sharktooth

that was... unexpected, however im happy they dont screw up...

Hi Sharky, you're the man!

Quote:

... on a sidenote eqm v3 ULR is obsolete since LCD displays. the reason is the "artificial" sharpening becomes an issue with high-contrast displays...

I persist; have done side by side 1 comparison with Jawor 1cd and ULR, some differences, but both are usable for that kind of extreme compressibility (ULR_VAQ is more compressible than Jawor_VAQ), seems VAQ rubs the artificial sharpening (even if ringing or artifacts still present).

I have not tested your UHR and EHR but Didée 6o9 seems to give excellent results with VAQ.

Did

henryho_hk · 17th February 2009, 13:01

Quote:

seems VAQ rubs the artificial sharpening

Perhaps VAQ saves enough bits to allow for lower quantizers.

I really hope that there will be XviD 1.2.2 which officially incorporates VAQ.

30th December 2008, 20:07	#1 \| Link
KikeG Registered User Join Date: Jun 2006 Posts: 8	VAQ and MPEG quantization matrix? I've read that VAQ should only be used with H.263 quantization. However I've tried using it with MPEG quantization and there's also a size advantage. On a 17 sec. clip of already encoded video of Monsters Inc., encoded at constant q=3, there's a 7'4% size advantage using H.2.63 and a 6,9% size advantage using a MPEG quantization matrix. As I see VAQ increases quantizers in areas with higher detail. I don't understand why this works any different with a MPEG cqm. Maybe it is because with VAQ quality can suffer more in case of MPEG cqm? Thanks.

31st December 2008, 04:39	#5 \| Link
Sharktooth Mr. Sandman Join Date: Sep 2003 Location: Haddonfield, IL Posts: 11,768	using a custom quantization matrix with VAQ may lead to unwanted results. but that depends on the matrix. __________________ MPEG-4 ASP Custom Matrices: EQM V1(old), EQM AutoGK Sharpmatrix (aka EQM V2), EQM V3HR (updated 01/10/2004), EQM V3LR, EQM V3ULR (updated 04/02/2005), EQM V3UHR (updated 17/12/2004) and EQM V3EHR (updated 05/10/2004) Info about my ASP matrices. MPEG-4 AVC Custom Matrices: EQM AVC-HR Info about my AVC matrices My x264 builds. Mooo!!!

16th February 2009, 14:48	#10 \| Link
Sharktooth Mr. Sandman Join Date: Sep 2003 Location: Haddonfield, IL Posts: 11,768	@henryho_hk: i no longer use xvid for my encodings since years. @KikeG: i said CUSTOM quantization matrices (MPEG matrix is not "custom"...). __________________ MPEG-4 ASP Custom Matrices: EQM V1(old), EQM AutoGK Sharpmatrix (aka EQM V2), EQM V3HR (updated 01/10/2004), EQM V3LR, EQM V3ULR (updated 04/02/2005), EQM V3UHR (updated 17/12/2004) and EQM V3EHR (updated 05/10/2004) Info about my ASP matrices. MPEG-4 AVC Custom Matrices: EQM AVC-HR Info about my AVC matrices My x264 builds. Mooo!!!

17th February 2009, 03:48	#12 \| Link
Sharktooth Mr. Sandman Join Date: Sep 2003 Location: Haddonfield, IL Posts: 11,768	that was... unexpected, however im happy they dont screw up... on a sidenote eqm v3 ULR is obsolete since LCD displays. the reason is the "artificial" sharpening becomes an issue with high-contrast displays. it is still valid if you have a CRT monitor/TV though... __________________ MPEG-4 ASP Custom Matrices: EQM V1(old), EQM AutoGK Sharpmatrix (aka EQM V2), EQM V3HR (updated 01/10/2004), EQM V3LR, EQM V3ULR (updated 04/02/2005), EQM V3UHR (updated 17/12/2004) and EQM V3EHR (updated 05/10/2004) Info about my ASP matrices. MPEG-4 AVC Custom Matrices: EQM AVC-HR Info about my AVC matrices My x264 builds. Mooo!!!

30th December 2008, 20:23	#3 \| Link
KikeG Registered User Join Date: Jun 2006 Posts: 8	I don't think I read wrong, however I didn't read it from you. Anyway, thanks for the clarification, it's what I wanted to know.

31st December 2008, 12:12	#7 \| Link
KikeG Registered User Join Date: Jun 2006 Posts: 8	In case of MPEG quantization matrix, what unwanted results?

31st December 2008, 14:39	#8 \| Link
Gromozeka Registered User Join Date: Jan 2007 Posts: 151	Mpeg matrix with VAQ give good result, but VAQ with ZSM matrix jawor-a - very bad result mpeg matrix q3 with VAQ very good too

16th February 2009, 11:48	#9 \| Link
henryho_hk Registered User Join Date: Mar 2004 Posts: 889	Sharktooth, would you please share your opinion or experience of your MPEG-4 ASP Custom Matrices with VAQ? H.263 is really a bit too soft.