Custom Matrices

[kilg0r3]

using the search, i found basically two different sets of custom matrices and experimented with them a bit. yet, since the corresponding threads have been abondoned for quite some time, i decided to create a new one.

to me it seems that the 'good matrix' and the 'better matrix' are those which can be used for most films. i am far from being familiar with the inner workings of a matrix, however and have not tested all of them extensively. for this reason i would really appreciate any input and even some new matrices to experiment with.

cheers to all of you.

[Teegedeck]

Hi,

it's been some time now, but I recall that I occasionally got strange glitches with the 'good' and 'better' matrices, so I stuck with H.263 for low bitrates. But I really liked the 'best' matrix; I used it to breath some artificial life into some sources that I had to denoise to death before encoding.

[kilg0r3]

right, i also remember that created the better matrix some grain like but static texture. If i remember right, the 'best matrix' was only for very low quants. That's why i never got round really test it.

Are there any people still designing new matrices?

[bilu]

@Teegedeck, kilg0r3

I've seen mixed opinions on Ultimate Custom Matrix (by Dali Lama) (http://forum.doom9.org/showthread.php?s=&threadid=33499&perpage=20&highlight=custom%20AND%20matrix&pagenumber=2).
Haven't tested yet, do you have an opinion?

I've tested "HVS Better" and "Best", and I liked "Better" more, colours seemed more natural. But at the time I did only CBR encoding at 860 Kbps, so I can't tell much about detail :rolleyes:

[kilg0r3]

could somebody point me to an understandable source explaining how quantization matrices work?

[droolian01]

Hi there

I have just started playing around with this feature. I've been trying to get a result that is somewhere between the sharpness of standard mpeg quant and the bluriness of h.263. I have very little knowledge of the technical aspects myself, but just looked at the numbers in Dali Lama's collection, combined with some tests.

Right - this probably won't line up but here goes -

intra frame
8 16 19 22 26 27 29 34

16 16 22 24 27 29 34 37

19 22 26 27 29 34 34 38

22 22 26 27 29 34 37 40

22 26 27 29 32 35 40 48

26 27 29 32 35 40 48 58

26 27 29 34 38 46 56 69

27 29 35 38 46 56 69 83


inter frame
16 16 19 22 26 27 29 34

16 16 22 24 27 29 34 37

19 22 26 27 29 34 34 38

22 22 26 27 29 34 37 40

22 26 27 29 32 35 40 48

26 27 29 32 35 40 48 58

26 27 29 34 38 46 56 69

27 29 35 38 46 56 69 83

basically these are identical except for the obligatory '8' in the intra frame.

So far it seems to produce quite detailed results in situations shere you haven't got lots of bitrate, but where 'low bitrate' is too soft. I have started using it for 352x193 100-125 mb encodes of buffy eps.

droolian

[AlphaDivxMovies]

I usually play a lot with the custom matrix but i always edit it to meet my objectives. For example i use this variation of the Best Hvs matrix for very saturated encodes. With this i managed to put The Others i 2cd easily otherwise i would only use 1 cd.
Quality wise the movies is better but not a lot better than the 1cd version, around 25% better not 100% better.
Please look at it and if you see some stupid mistake tell me!
Hope it is of some use.

[kilg0r3]

@Alpha...

Thanks for your matrix. Unfortunately i haven't been able to test it cause my dvd drive is dead :(. Anybody else here who could make some tests?

IMO this matrix issue deserves much more attention but seems to be a difficult area. Again, could anybody point me to an understandable source of info about how these matrices work?