CCE Matricies... which is best for Anime material?

[Por_Que2k]

I've been messing with the CCE matricies for a few days and had some results, but not what I was hoping for. Do any of you advanced encoders have some matricies that are good for Anime?

[homerpez]

I'm giving this one a >BUMP< because I'm very interested in an answer to this...

[KoalaBear]

Quantization matrices are used by an MPEG encoder to achieve bitrate control. You really shouldn't mess with them unless you know exactly what you're doing, which you obviously don't or you wouldn't be asking this question in the first place.

I don't mean to be harsh, but you're barking up the wrong tree here. You'd get more practical answers if you asked about techniques for handling animation in general.

[Por_Que2k]

Ok, I was just wondering about the matricies. I know how to handle animation in the other fields of encoding.

[KoalaBear]

Then may I ask what led you to think quantization matrices were a promising target for optimization?

I don't do animation, but based on general MPEG principles I'd say shorter GOPs with fewer B-frames were the way to go. What is it about the quantization table that makes it look so sexy to people?

[This is an honest question and not a troll. I'm genuinely curious about the answer.]

[kinneera]

It can have a substantial impact on the encoding quality of difficult material for MPEG (such as animation), as well as low bitrates. TMPGEnc has a CG/Animation quantization matrix, which would probably work well for anime. This quantization matrix can also be patched in to CCE, I'll leave it to you to figure out how. (In any case, my experience has indicated that TMPGEnc does in fact perform better with animation, for whatever reason).

<font size=-1>[ This Message was edited by: kinneera on 2002-01-17 14:55:12 ]</font>

[homerpez]

KoalaBear, I think your comment was severely out of line. I have this vision of you as having a huge head that makes you fall down a lot.

It's perfectly fine to ask questions, and what's more this was a valid question to be asking. Lay off.

kinneera, I'd like to thank you for your MATURE response to this valid question.

[maynardjames]

Buddy got BURNED!!!!


I am really sorry guys, but when i see a burn...............



Oh yeah and i encode a lot of anime too so i am allways looking for ways to make it look better in even the slightest bit so........

<font size=-1>[ This Message was edited by: maynardjames on 2002-01-17 15:36:06 ]</font>

[KoalaBear]

Homerpez: In what way (do you think I'm "out of line," that is)? Because frankly I'm just trying to save you some trouble by pointing out that if you can't even define the term, you're not in a good position to manipulate the concept to your advantage.

Kinneera: While several MPEG encoders (Panasonic for example) offer alternative quantization matrices nominally optimized for particular encoding tasks, (1) they are the product of commercial R&D versus mindless tweaking, (2) they are not guaranteed to deliver the results you think they might, and (3) in a VBR environment the encoder dynamically optimizes those tables on a per-GOP basis if necessary as a result of the VBR algorithm itself.

One of the risks of altering these matrices without a clue as to what you're doing is that you can force an underflow or overflow of the MPEG frame buffer during decode. The values that make up the matrix aren't chosen at random, they're tuned to the characteristics of the particular encoder and often take into account performance characteristics that can't be predicted externally.

[kinneera]

There's really no reason for anyone to get testy, although I'm in the party that agrees with the idea that merely asking about it is perfectly fine.

@KoalaBear - this answer was useful. Why not just post it to begin with, rather than making vaguely insulting remarks like "What is it about the quantization table that makes it look so sexy to people?"?

In any case, I'm aware that all such customized quantization matrices are a result of R&D by people who understand them. That doesn't mean there isn't a benefit to bringing them to people's attention where they exist.
A lot of people use CBR, so just because VBR will modify them on the fly doesn't mean that custom matrices aren't still useful. I would imagine that they still have a pervasive effect as the starting point even in VBR encoding.

[KoalaBear]

I think it's human nature to hope that the piece of the puzzle we don't understand is the one that contains the key to the solution. People who aren't hip to what GOPs are about, for example, are the ones most convinced that they can do better than the default.

[homerpez]

Quantization matrices are used by an MPEG encoder to achieve bitrate control. You really shouldn't mess with them unless you know exactly what you're doing, which you obviously don't or you wouldn't be asking this question in the first place.

I guess the question is, if you were looking for help, and someone made sure to tell you, in no uncertain terms, that you "obviously don't know what you're doing", how would you feel about that?

I'm only saying it doesn't add anything beneficial to the discussions, forum, solving problems if you add these little digs... it doesn't make for pleasant reading...

I like more the last post you made, it certainly sounds better than the first. It's fine to make points, but make them in such a way that doesn't have the sole purpose in mind if inflating your ego.

[vitualis]

I think that there is just misunderstanding here... I don't read KoalaBear being deliberately offensive here and I think we should leave it at that.

Back to the question of quantization matrices... I can honestly say I have no idea how to tweak them and would probably leave them alone as you can definitely do more harm than good.

With the alternate matrices available in some encoders (e.g., Panasonic), even Panasonic in their help file say that the alternative settings (e.g., CG) probably will not yield better results than their default one.

If anyone is a particular "expert" in this, I'm interested in knowing the answer too.

Regards.

[homerpez]

Sorry, virtualis aka. Michael. I just get sick of certain comments which would bother me if I was seeking help here.

Koala, no hard feelings. Hope you know where I was coming from.


---------------------------------------------------------------------------------


Back to topic, I too can testify that the TMPGEnc matrices that are supposed to be for CG/Animation really didn't help me all that much. In fact, if I remember the test correctly, the CG/Ani matrix made it even worse.

I too suspect that other factors make more of a difference, I've noticed personally that some animation inherently craves more bitrate. But who knows? Maybe this is some sort of "Holy Grail"?...