details details

[ammer]

Hey is there a way to add 3.11 option to xvid to get the extra details?

[Selur]

"Hey is there a way to add 3.11 option to xvid to get the extra details?"
Hmm,.. and of which options are you speaking exactly of?

Hopefully it's nothing one would need the source code of M$-mpegv3 for, because as far as I remember only M$ got this, and it's not open source,..

Cu Selur

[Neo Neko]

Originally posted by ammer
Hey is there a way to add 3.11 option to xvid to get the extra details?

Yes of what are you speaking? Even Divx 4.0 has all of the same controlls as Divx 3 and then some? And Xvid has Divx 3, 4, and 5 all beat for the most number of settings. Are you speaking of Nandub? If so Keopi and -h have stated many times that the options from nandub that are not implimented in Xvid would have been useless to Xvid anyways. Only the aplicable ones made it through.

[Teegedeck]

Originally posted by ammer
Hey is there a way to add 3.11 option to xvid to get the extra details?

HEy this kinda questions really make my day! :)

[gldblade]

Maybe ammer was thinking about adding the option of using DivX 3.11 in XviD. IF this does happen to be the case, ammer is a little confused.

[OUTPinged_]

he probably was referring to that h.23-something vs MPEG quantizers setting topic that was here long ago.

ammer, just browse through archive.

[ammer]

It'd be good to have some more extra details but I was just kidding about the 3.11(divx3.11) thing (maybe something of that type though maybe). It was for all those asking for the divx3.11 details not in XviD. Seems to me that using divx3.11 to get those extra details would make more sense in that case anyhow.

[-h]

I still don't know what you're asking for.

-h

[Koepi]

I'm with -h here...

:confused:

[Selur]

I beleave, that after reading doom9s guides (and may be doing some tests of his own), he realized that sometimes DivX3.11;) is sharper/more detailed that DivX3.11;). So he's asking if it is know why this is so and if it would be possible to 'fix' this 'handicap' of Xvid,...

Cu Selur

[skyline]

@ ammer

Yes there is a way!
When you go to the compression options in vdub, select "XVID" compressor. When dialogue window appears, press alt-F4.
This should bring you the options you are looking for.
Note: you may have to first activate the options BEFORE opening virtualdub (if you havent already done so) by going to start-->run and typing "nul/nul" (win95/98)
Hope this helps :)

ps. Don't follow the above advise.

[Hanty]

We are still guessing if this is what he's really talking sbout though. The next question will be "Can you put DRF levels in XviD?"

[Neo Neko]

Several people have asked that already I believe. Only to be slapped down for outright sillyness. :sly:

[Hanty]

That was sorta the point.

[-h]

There's some "thresholding" performed in the quantization routines for H.263, whereby extremely small coefficients are zeroed instead of stored, thus some level of detail is being discarded. I can't imagine they'd be causing this however.. I'll try a test build with the thresholding disabled.

-h

[rui]

Maybe the way to go isn't that one.
H.263 smoothes the image a little more than MPEG, correct? This can be usefull for those making 1 cd rip's of dificult movies. So, why mess with it?
And for those, like me, who like things more to the "sharper" side (lesser lost of details, right?), then use MPEG quantization.

Maybe the solution for getting those details would be in a costum matrix builted with "sharp-details" in mind.
I remember having read, when costum matrix first appeared http://forum.doom9.org/showthread.php?s=&threadid=18938&highlight=costum+matrix
,that -h would make available to us a matrix that would give us a middle ground between H.263 and MPEG, that we would be hable to load using the LOAD button.
Maybe we could have another costum matrix that would be even more sharper than normal MPEG?
But i must admit that don't i have a clue of how to do that matrix :rolleyes:

[-h]

H.263 smoothes the image a little more than MPEG, correct? This can be usefull for those making 1 cd rip's of dificult movies. So, why mess with it?

DivX 3.11 uses H.263 quantization, thus it should be possible to preserve the same level of detail with XviD's H.263 quantization. It might be useful as an option, anyway :)

And for those, like me, who like things more to the "sharper" side (lesser lost of details, right?), then use MPEG quantization.

If you blow up the final result though, DivX 3.11 still retains more details. MPEG introduces noise, which manages to trick the eye into perceiving more detail.

Maybe we could have another costum matrix that would be even more sharper than normal MPEG?

I'll have a play. Might make file sizes a tad larger though.

-h

[rui]

Originally posted by -h
DivX 3.11 uses H.263 quantization, thus it should be possible to preserve the same level of detail with XviD's H.263 quantization. It might be useful as an option, anyway :)

Yes, but we know that SBC divx3.11 rips also contain more blocks than xvid. So divx311 is trading the blocks for more details, right?
If xvid starts doing the same, wouldn't we have more blocks when using that option you mentioned?

Originally posted by -h
If you blow up the final result though, DivX 3.11 still retains more details. MPEG introduces noise, which manages to trick the eye into perceiving more detail.:)

MPEG quantization has the effect of some kind of pre processing noise filter? Or am i way off? :)

Originally posted by -h
Maybe we could have another costum matrix that would be even more sharper than normal MPEG?

I'll have a play. Might make file sizes a tad larger though.

-h

As it stands know, the default costum matrix values give us less, equal or more sharper results than standard MPEG quantization?
I know, i know, i should test and compare :) but what results should i expect from those default matrix values?

[-h]

Yes, but we know that SBC divx3.11 rips also contain more blocks than xvid. So divx311 is trading the blocks for more details, right?
If xvid starts doing the same, wouldn't we have more blocks when using that option you mentioned?

Perhaps. I'm not entirely sure.

MPEG quantization has the effect of some kind of pre processing noise filter? Or am i way off? :)

Nah, it's just the way MPEG quantization works - it adds "invented" information, which in effect is noise.

As it stands know, the default costum matrix values give us less, equal or more sharper results than standard MPEG quantization?
I know, i know, i should test and compare :) but what results should i expect from those default matrix values?

The default matrix is the matrix used by MPEG quantization normally. If you leave it as is, you'll get the same results as just picking MPEG instead of MPEG Custom.

-h

[rui]

OK, boss. :D

Thanks for being always available to help us understand this "things" a little better ;)

Be sure that your dedication is noticed and VERY apreciated. :)

[-h]

Well I ran a quick test, and I'm not sure what's going on now.

I've attached a picture which displays (from left to right):

Original
DivX 5 (H.263)
DivX 5 (MPEG, enabled through registry)
DivX 3.11 (H.263)
XviD CVS (H.263)
XviD CVS (MPEG)
All are at quant=2, and my initial impressions are:

DivX 5 H.263: Blurs the most of any
DivX 5 MPEG: I'd say it's buggy, too many details are being invented
DivX 3.11: A bit sharper than DX5H263, evidence of poor motion estimation in collar (middle of screen, 75% from top, sharp corner appearing on what should be smooth line along collar)
XviD H.263: About equal with DivX 3.11 - skin and collar are a bit better, grey jacket is worse
XviD MPEG: Closest to the original, IMHO
But anyway, from the tests I just did I'm not sure if it's worth looking into.

-h

[rui]

Great work!!
Agree. Xvid MPEG looks great!

But this is with quant=2. With a real movie rip, quants not always are equal to 2. If one goes for 1 cd, almost never is.
But that's why we have modulated quantization, right? :D

Which brings me to the next question: i hope you guys aren't forgeting to solve the modulated quantization and MV hints bug.

And, has we are at it, i remember having read in the xvid.org forum an idea about the modulated quantization not being so restric. I mean, with quants=2,3 we have MPEG quantization. With quants >3 we have H.263 quantization. The guy that brought this up was questioning if this isn't a too big change, he was sugesting to have some kind of middle ground. When i read it, it sounded right.
Could something like this be implemented (in spite that i don't know what one would understand by middle ground here)?

[ammer]

Hmm nice pic line-up. Well done. If not clear, i'm just telling others to encode with divx3.11 for details it has. Point in case pointed. However, something of an option that could be a useful might be one that could be for improved encoding for ultra low coding bitrates (maybe a whole new coding scheme or matrix and maybe non-iso compliant) or one for minimal or higher bitrates for more closer to original quality encoding. Sort of a more perfectionized of quality. of a "higher qualitative" encoding...

[Neo Neko]

That would present a bit of a problem. At this point the whole Xvid project and all those devoted to it are aimed at developing a standards compliant MPEG4 implementation. Which means no non-ISO stuff will be included.

[Dali Lama]

Ammer, check out Real Video 9. They have a free command line encoding tool from their website. realnetworks.com If you are interested in a non-standards compliant codec, this should be a nice alternative to Xvid (the alternative MPEG-4 codec :) )

Real Video 9 gets rid of a lot of the anoying problems present in MPEG-4 at low bitrates (ie: below 1000kbs). I posted in the General Section under "Real Video 9" I posted two trial clips I encoded. You might want to check it out before you encode, to see what type of quality you can achieve.

good luck,

Dali

[ammer]

Yeah, but it doesn't quite have that open source, non-money-grubbing-corporate, feel free to innovate when ever you want to feeling to it. But i've tried Real Video encoding, it sort of locks you into their tools, encoding and decoding wise. Though it encodes some pretty cool mini video streams, not that the other encoders don't. Regardless of their coorporate greediness or proprietary techno i'll have a look at their latest encoder and see if they've pushed encoding to the next decimal place.