Standard (as in CCE) matrix?

[MiCK76]

hi there! is anybody using the standard matrix (the same "standard" used in CCE and given with the koepi binaries) instead of the h263 or mpeg matrices? isn't that better (or a good compromise) than h263 (which if i understood well seems to have the same factor =32 for all frequencies) and mpeg (which has the same intra matrix like the standard i'm using but the inter one with all factors =16)?
i tried it for a few encodes with Xvid RC3 with bitrates 900 to 1300 kbps (always aiming for a Bits/pixel*frame 0.20 to 0.30, anyway given a 60 to 70% compressibility test) and seems pretty good.
less blurry/soft pictures than h263, almost detailed as mpeg.

[MiCK76]

ok i just tried experimenting with custom matrices, now discovering 6of9-hvs or hvs-best matrices to test soon...
just a question as a newbie in this matrices argument:
for an encode like the one described above (900 to 1300 kbps etc..) what's best? i read 6of9 is for really high bitrates (i assume 2mbps at least) so is hvs-best more suitable for my encodes?
just need a few suggestion to start my journey in the right way,thx.

[EthanoliX]

According to what I heard, hvs-best seems to be suitable for your purpose. But take care, which matrix works best, does not only depend on bitrate and resolution, but also on your source.

[MiCK76]

ok thx EthanoliX, assume i'm always using DVD sources, so is this still correct?

[EthanoliX]

Originally posted by MiCK76
ok thx EthanoliX, assume i'm always using DVD sources, so is this still correct?

Not exactly. I was more talking about the properties of your source: Is it dark/bright, sharp/smooth, grainy, noisy, high/low motion ...

Depending on how difficult a source is to process for the codec, the higher the aquired bitrate for getting the same quality or the lower quality using the same bitrate.

Best would be to run a few tests on different sources, using different matices and bitrates and see for yourself, which looks best.

Unfortunately there is no rule saying "Exactly this matrix for that kind of source". ;)

[Chainmax]

I also have a question: how does hvs_best play with Trellis and/or QPel? Also, would you use it on a cartoon rip if the source is properly denoised and smoothed?

[EthanoliX]

Originally posted by Chainmax
I also have a question: how does hvs_best play with Trellis and/or QPel? Also, would you use it on a cartoon rip if the source is properly denoised and smoothed?

1. AFAIK yes.

2. Sorry, I have no idea. I don't do cartoon rips.

[MiCK76]

Originally posted by EthanoliX
Not exactly. I was more talking about the properties of your source: Is it dark/bright, sharp/smooth, grainy, noisy, high/low motion ...

Depending on how difficult a source is to process for the codec, the higher the aquired bitrate for getting the same quality or the lower quality using the same bitrate.


well thx again, but that's why i run my compressibility test, don't i? ;) i aim always to about 70% of my full quality test, so the bitrate i reported was only indicative (or better it was the result of my 4 last encodes). but what i mean, aiming for the same "relative" quality (say 65to75% of comp test), the bitrate is given for any source this way, so i can stick with the preferred matrix (once found, seems like hvs-best is my first candidate for now) for every (DVD) source. correct me if i'm wrong, these are only my 2 cents, and surely i don't pretend to be any master..it's just the silly idea i got on mind not to waste many hours to run different tests for every source i got...

anyway, i take the opportunity also to say a BIG THX to Xvid team.your work is awesome, you rule men! maximum respect.

[MiCK76]

Originally posted by Chainmax
I also have a question: how does hvs_best play with Trellis and/or QPel? Also, would you use it on a cartoon rip if the source is properly denoised and smoothed?

just tested it with qpel, gmc , (turbo mode on), trellis quant...
seems pretty good man, try it yourself if you don't trust!;)
for cartoons i can't help you...don't really know, never tested it

[Soulhunter]

Originally posted by Chainmax
Also, would you use it on a cartoon rip if the source is properly denoised and smoothed? AFAIK it copes good with such stuff... ;)


Bye

[Arcon]

Originally posted by MiCK76
now discovering 6of9-hvs or hvs-best matrices
i've searched for 6of9 and found a pretty old thread where somebody linked to a file with this matrix but i've got no idea if this is the final/actual version (or if there was more than 1 version :) ). so if you still got the links or values for 6of9 and hvs-best i'd appreciate it if you could post them here.

btw: since many people seem to use them for high bitrate encodes i wonder if they will get included as goodies in xvid 1.0?

[Soulhunter]

Originally posted by Arcon
So if you still got the links or values for 6of9 and hvs-best i'd appreciate it if you could post them here.
Here... (http://forum.doom9.org/showthread.php?s=&threadid=72211&perpage=20&pagenumber=2) ;)

[Arcon]

Originally posted by Soulhunter
Here... (http://forum.doom9.org/showthread.php?s=&threadid=72211&perpage=20&pagenumber=2) ;)
hm, that wasn't that far away. thanks :)

[P0l1m0rph1c]

XviD already brings some quant matrixes for you to play with :) . Just go to your XviD instalation folder.

LiGH has a very complete quant matrix collection. I have the file mirrored here (http://moolabs.dcc.online.pt/downloads/codecs/qmatrix.zip) .

EDIT: crap, too late...