Bilinear, Bicubic & noise filter!! ?þ|

[SirTomahawk]

Hello,
problem: before I finally let encode a film, I test a lot with the filters. For example: (5% compressibility test)
1. bilinear -> 64.0%
2. neutral bicubic -> 54.8%
3. sharp bicubic -> 52.3%

As you can see, bilinear seems to rule at this value. I take this and continue testing:
1. bilinear + "low noise filter" -> 70.8%
2. bilinear + "low noise filter" +
"noise filter before resizing" -> 71.8%

Can it be? What is best? I do some test encodings and it seems, that this percentage cannot be true, because the video with this 71.8%looks worse as the sharp bicubic.

I read the help in gordian, but there is no explaination. When I have to prefer "bilinear" filter?

Please help and explain, T.

[Krack]

:p
In compressibility test there is "compressibility".
So I think that is just to see if you can reach the size of your movie you choose before.
Besides quality depend of the resolution of your movie and choosing sharp bicubic instead of soft bicubic can make the video more blocky.
for the noise filters I just know that it makes the images less floating so I make the most of times sharp bicubic with little noise filter on 2 cds. :p

[SirTomahawk]

Hello,
yes Krack, I know. But I wonder, that the filters can improve the compressibility so much! Noise filters are ok (only at low). Over low you could get such like this:
Link (http://www.tomahawk.miesto.sk)

Now I know, that noise filter kick some details and so it raise the compressibility.
But what's the exact difference in bilinear and bicubic? I know that bicubic is best, because the sample range is bigger as bilinear. But sometimes I read, that you should take bilinear. Why?

T.

[Krack]

yeh very good examples. :D
But what filter have you used? :confused: only for the little noise? :confused:
I don't know what make really these resize filters but I have try once the bilinear and it was very bad quality. :(
:cool: I think that we have to use it only for a very hard compression (under 600 kbis/s): it allows to have less block.
In fact these filters are in increase order of cripness but blocky image too. So if you have a good Kbit/s you can take sharp bicubic.
I wonder if this check compressibility is just for divx4 because I have always reach the perfect size with divx3.11a whenever I don't respect the compressibility check. But then I have more blocks.
If you have read the very good test of Doom9 you can see that divx4 have fast no blocks but have size problems. :cool: