1PASS CCE quantizer matrices

[Delta2]

Hi,

Which is the best quantizer matrix for 1 PASS VBR encoding with CCE to produce the best quality DVD (MPEG2) ?

I allways use standard, but I want to improve quality ... does this option of "quantizer matrices" make any difference on DVD quality ?

thanks

[manono]

Yes, the matrix used can and will make a big difference in the quality of your encodes. And if the source is good (another DVD, for example), then CCE's default Standard Matrix is a very poor choice. Have a look at the matrices used by your favorite Hollywood DVDs. Do you think any of them use the Standard Matrix? Not on your life.

1-pass VBR, as in CQ? The fact that you're running one pass doesn't mean you should use a special matrix for it. You use a matrix that fits the file size you have available, the compressibility of the source, and the Q-Factor (ave. quant) you like to achieve. Sort of. This is fairly complicated, and asking for a matrix for a 1-pass VBR encode shows you don't quite get it yet. But you're on the right track, as you're at least aware of the issue now. Just because you might want to use a high bitrate, detail retaining matrix, doesn't mean you can all the time.

My suggestion is to collect matrices from commercial DVDs, type them into CCE or paste them into the .ecl file, and start testing.

[Delta2]

My suggestion is to collect matrices from commercial DVDs, type them into CCE or paste them into the .ecl file, and start testing.

thanks manolo, but how can I collect matrices from commercial DVDs ? How can I rip that information from a DVD ? can you say me more please ?

[manono]

The best way to get the quantization matrix (matrices) used in a video is to open the Vob(s) in DGIndex, go Options->Log Quant Matrices, and then File->Save Project. When done, in addition to the D2V file and the audio stream(s), you'll also have a Movie.quants.txt file. Look inside there. Make sure you have an up-to-date version of DGIndex. The newest one also tells you which frames the various matrices are used. The older versions can't Log Quant Matrices at all. To whet your appetite, here's a very nice high bitrate matrix taken from a commercial DVD:

08 08 09 09 10 10 11 11
08 09 09 10 10 11 11 12
09 09 10 10 11 11 12 12
09 10 10 11 11 12 13 13
10 10 11 11 12 13 13 14
10 11 11 12 13 13 14 15
11 11 12 13 13 14 15 15
11 12 12 13 14 15 15 16

08 08 09 09 10 10 11 11
08 09 09 10 10 11 11 12
09 09 10 10 11 11 12 12
09 10 10 11 11 12 13 13
10 10 11 11 12 13 13 14
10 11 11 12 13 13 14 15
11 11 12 13 13 14 15 15
11 12 12 13 14 15 15 16

[Delta2]

uau...I tested and the quality...uau

manono...what is considered as High Bitrates ? values greater than 2000kbs ?

As far as I understood I must use a different matrix for low bitrates...isn't it ? do you have any

[guns1inger]

For DVD, high bitrates are up around 7500 or more. I would consider 3000 - 3500 in the low range

[Delta2]

then...I must not use a High Bitrate Matrix for encoding a movie with bitrates between 2000 and 3000 ? isn't it ? I must use a slower bitrate, medium bitrate and a higher bitrate matrix ? or can I use the same matrix without loss of quality ?

For High compression or more encoding passes the bitrate are less than 4500

this is interesting

[manono]

There are several factors you have to consider when deciding on the matrix you're going to use. One would be the matrix of the original encode. If, for example, it used the Standard matrix, so much detail has already been removed that no matter how good the matrix you decide to use, you won't get that detail, that sharpness, back again. So, you'll have your best luck being able to use a high bitrate matrix such as that one above, if an equally good matrix was used for the source.

Another factor is how much you'll be compressing from the original. In general, you'll be able to use a better matrix if you want your reencode to be 90% of the original size, than if it's only 50% of the original size.

And one very important factor is the amount of overhead that the original allows for reencoding, overhead being measured by the Q-Scale, or average quant of the original. This is something that DVD Shrink users don't seem to understand. If the average quant is high to begin with, you have no overhead for transcoding. Shrink and other 1-clickers keep the same matrix, and they will turn to crap very quickly. That's why all discussions about what percentage you can use with DVD Shrink and still have it look good are pretty much worthless, and only participated in by the clueless. When you're reencoding, you have the option of using a lower bitrate matrix (although most people use the default Standard Matrix because they don't know any better), so you can compress it quite a bit, and it will still look good. But this concept of overhead is important. If the average quant of the original was 2, you can compress the heck out of it and it will still look good. If it was originally 10, then you might be in trouble.

And although I call it a high bitrate matrix, or a lower bitrate matrix, bitrate doesn't really have a whole lot to do with it. A much more important figure is the Q-Scale, or Q-Factor, or average quant.

And if you want to use good matrices, but don't want to do the work of saving them and studying them, then one very good alternative is just to use the Standard Matrix with AQ turned on in CCE (Advanced screen, with the "Disable Adaptive Q-Matrix Switching" box unchecked). Then you'll have a mix of 3 matrices, the Standard Matrix, the Standard Matrix with the numbers all halved, and the Standard Matrix with the numbers all quartered. This Standard Matrix with AQ (Adaptive Quants) turned on is found on quite a few retail DVDs these days.