Temporal smoothing, yes or no?

[da franksta]

Should I or shouldn't I?
Is there a catch?

TIA

[chainsaw135]

Well there is always a catch:)...really there are alot of reasons 2 use and i'm sure just as many not to use temporal smooth. Some I can think of are...

1.Is my bitrate 2 low and maybe I need 2 smooth the images out?
2.Do I wanna take a 40% hit on encode speed?
3.Depends on the source also
4.Am I using cbr,vbr 1, or multipass.


I know I havent covered even close 2 all the reasons of "when you might or might not use temporal smoother" ..but those are some of the reasons I can think of off the top of my head

[DDogg]

My two bits is:

1> TS 2,1 On with any one pass mode
2> TS 2,1 On with multipass IF bitrate is under 1600

Also, in the new version, Bilinear is available in the Framegrabber tab. I would suggest Bilinear for lower bitrate encodes as Bilinear compresses better.

[Northpack]

Hi DDdog,

after reading a lot of threads here about the usability of the TemporalSmooth-Filter, one question still remains to me: everyone seems to use a strength of 2 in conjunction with a radius of 1. So why man should not use a higher value? As the name implies, I assume that the filter would be more powerfull with more comparison material (e.g more frames) to work with? I've played arround with the settings in order to figuring out what different values in both parameters effect - if strength is set too high, the results get realy awfull (kinda spooky ;)), so I would never set them about the value of 3 for the purpose of svcd encoding. On the other hand, as I've changed the radius, the effect seemed not quite as clear to me, besides that the playback speed slowed *way* down! But if we don't care about lousy encoding speed - does it *visually* make sense to use a higher radius? Or otherwise: could it do any kind of *harm* to the image-quality (like strength does easily)?

Thanks,
Northpack

[DDogg]

I am kinda a "first do no harm" kind of guy :)

What we are trying to do with TS 2,1 is help compressability, rather than true filtering although one could argue that they are the same thing.

We want to get the adjacent frames as close to each other as we can so excessive bitrate is not needed (wasted), but with the least impact to the actual visual. This will free up bitrate that can be better put to work storing the video rather than noise.

Hope that makes a little sense. Ah, I found an old thread that may be of some interest. Take a look. http://rilanparty.com/vbb/showthread.php?s=&threadid=7764

[Northpack]

Me too, DDog, me to! ;)

I already read this thread. What Dividee says is undoubtly right; a strength of 2 seems so suite perfectly for improving cempressibility in most cases. But are there any downsides, besides speed, to use a higher value on radius? After all what i've read about this filter, I assume that it should work even better, if it has more frames to work with?

Thanks,
Northpack

PS. I recently encoded Starship Troopers (16:9, 124min) to 2x800MB and I'm totaly blown away by the results! On the TV (it's not the best screen however) it's hardly to differ from the original DVD-source! I used TemporalSmoother(2,2) in conjunction with a Bicubic of 0-0.5, IQP 20, the CCE-internal Noise-Filter on the Value of 4, and heavy macroblock optimization (24 pixel black-bars to the left and the right - no screen will ever show this range!). The biggest improvement however, brought the usage of an optimized SCVD quantisation Matrice (Andreas SVCD!) - i did the same encoding before, using the CCE standard matrice. What a difference!

[mOE]

@Northpack

Where can I find this or other SVCD optimized quantisation matrices for CCE?

[DDogg]

Northpack, it sounds like you have done some good research and testing on this subject. How about a separate post fleshing it out some more (with links, how to, etc.)?

Maybe we can stick it in the advanced forum for reference.

As for: "But are there any downsides, besides speed, to use a higher value on radius?". I really don't know the answer. I would suggest you ask Dividee or Steady in the AviSynth forum. They really know this stuff cold. Mozart is good on the subject also.

[Spawn]

I personally use TS(1,2), I find a strength of 2 far too much, objects get a kinda weird appearance. When I encoded Enemy at the gate using a strength of 2, on a regular TV but on my HDTV everyone looked like there heads were polished. Since then I stick with 1 and it looks perfect.



Northpack: I would also be interested in these SVCD matrices, please post them.


-SPaWN

[orbit-r]

hi
can you please post a link to the matrice
cause i wanna try it

thanxs
orbit-r

[bradnopit]

link for the optimized matrix from ANDREAS (http://www.dvd-svcd-forum.de/)

Have fun testing, it gave me some good results

[wfiguero]

Are this the Andreas quantisation Matrice ?



Quantize matrix

Einstellungen von Andreas mit 99er-Werten und NEUE Anordnung

Intra
_________________________________
08 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

NonIntra
__________________________________________
16 17 18 19 20 21 22 99
17 18 19 20 21 22 23 99
18 19 20 21 22 23 24 99
19 20 21 22 23 24 26 99
20 21 22 23 25 26 99 99
21 22 23 24 26 99 99 99
22 23 24 26 99 99 99 99
23 24 25 99 99 99 99 99
____________________________________

Soften block noice

Intra block........................ 5-10
Non-intra-block.................... 4-8, höherer Wert kann Pumpen verursachen.

Weiterhin macht es Sinn, unter USE FLOTING POINT DCT
( früher iDCT) einen Haken zu setzen. (Andreas)

[mOE]

Yes, it is...

[orbit-r]

hi
tried the andreas matrice and it looked very cool..

i`m new to the matrice thing, but I´m going to try
the non intra settings from àndreas
and the intra settings from the low-bitrate-matrice,cause somewhere i read the intra settings are for low bitrate (since the intra settings from andreas are the standard settings of cce why not?)

what do you think about this?

thanxs orbit-r




:rolleyes:

[Northpack]

For a bitrate about 2000, use the matrice posted by wfiguero. For a bitrate of 2000 and less use this one:

Intra:

8,16,19,22,26,27,29,99,
16,16,22,24,27,29,34,99,
19,22,26,27,29,34,34,99,
22,22,26,27,29,34,37,99,
22,26,27,29,32,35,40,99,
26,27,29,32,35,40,99,99,
26,27,29,34,38,99,99,99,
27,29,35,38,99,99,99,99

Non-Intra:

16,17,18,19,26,21,22,99,
17,18,19,20,27,22,23,99,
18,19,20,21,29,23,24,99,
19,20,21,23,29,24,26,99,
20,21,22,23,32,26,99,99,
21,22,23,24,26,99,99,99,
22,23,24,26,99,99,99,99,
23,24,25,99,99,99,99,99

Don't forget to mux the matrice with bbMPEG into the header of the mpeg-stream afterwards. CCE 2.5 originally doesn't support the use of custom matrices, thus it writes the standart matrice into the .mpv header and the decoding quality suffers because the decoder hasn't got the adjusted matrice to work with!

Greets,
Northpack

[Northpack]

Or, alternatively use one of the following two matrices:

Angel SVCD Low
8,13,15,19,19,23,23,29,
13,13,17,19,22,23,27,29,
15,17,19,22,23,27,29,35,
17,21,22,23,27,29,35,37,
21,21,23,27,31,33,37,46,
21,23,27,31,33,37,46,58,
25,25,31,33,37,40,54,69,
25,31,33,37,40,54,69,78

8,11,13,15,15,19,19,24,
11,11,15,15,19,21,24,25,
13,15,17,19,21,24,25,34,
13,17,19,21,23,26,34,99,
17,19,21,23,26,32,99,99,
19,21,23,26,32,99,99,99,
22,22,28,30,99,99,99,99,
22,28,30,99,99,99,99,99

Angel SVCD Ultralow
8,13,15,19,19,23,23,29,
13,13,17,19,22,23,27,29,
15,17,19,22,23,27,29,35,
17,21,22,23,27,29,35,37,
21,21,23,27,31,33,37,99,
21,25,27,31,33,37,99,99,
25,25,31,33,37,99,99,99,
25,31,33,37,99,99,99,99

8,11,13,15,15,19,19,24,
11,11,15,15,19,21,24,25,
13,15,17,19,21,24,25,34,
13,17,19,21,23,26,34,99,
17,19,21,23,26,32,99,99,
17,21,23,26,32,99,99,99,
22,22,28,30,99,99,99,99,
22,28,30,99,99,99,99,99

These are the infamous Angel-SVCD matrices. I can't tell wether they are better / worse than the Andreas matrices. Both are optimized for SVCD-Encoding of progressive material. Angel's matrices were originally developed for CCE, Andreas for TMPGenc but it seems that both perform as well on the each other Encoder. Perhaps you should give them both a try...

Greets,
Northpack

[tyee]

Hi Northpack
Is it possible to do this matrix muxing for MPEG2 DVD encoding too?

Tyee

[orbit-r]

hi tyee,
i think its possible to mux the matrices into dvd mpeg2(vob`s)
all you have to do is
bbmpeg>>>>start encoding>>>>settings .. in settings choose `programm settings`and check it to dvd.... and than load the matrices(you like)
under the `in and output tab`(and shurly your mpv and audio streams..)

i think this will work.. but in the minidvd guide, doom reported some problems when muxed with bbmpeg

so far,i think to mux the matices into the stream for dvdmpeg only makes sense if you doing minidvd (or dvd with lowering the bitrate by a encoder).cause a ifoedit copy will have the original matrices(i believe) im not (yet)into the dvdcopy thing....;)

gruss
orbit-r