Denoising making bigger filesize?

[koolcracker]

I have been testing out some denoising filters on some DVDs I have, but everytime I do, I am consitantly getting larger filesize WITH the denoisers than without. I have tried Deen, Unfilter and Fluxsmooth, all with the same results. I use a basic script, maybe there is a problem with it:

MPEG2source("C:\Rips\BatmanBegins.d2v",cpu=4)
ColorMatrix()
FluxSmoothST()
BicubicResize(640, 480)

Can someone here tell me what is going on? Is there a solution? or is this jsut something that will happen? Thanks for the help.

[foxyshadis]

By comparison, you mean comparing against this:

MPEG2source("C:\Rips\BatmanBegins.d2v",cpu=4)
ColorMatrix()
BicubicResize(640, 480)

right? (Also, unfilter doesn't really count as a denoiser, at best it's just a very mild median cut, for which removegrain(1) is better suited; unfilter is more for a removal of mild oversharpening or blurring.) Also, what codecs? Oddly, cleaned-up lossless is usually larger than the original, and I can't explain why, but dct-based codecs are almost always smaller with quality encoding. More details in general would help; for instance, it could be that a custom matrix reverses the normal behavior.

[Chainmax]

Why aren't you using info=3 in MPEG2Source along with hints=true in ColorMatrix?

[koolcracker]

I am using XviD, but I am also using a custom matrix, and I didn't know they could reverse things like that. What exactly does the info=3 and colormatrix hints do? im not quite clear on that, but does it help them to work together? Thanks for the help. Ill report back.

[Mug Funky]

when the noise level is high enough (higher than most DVDs mind you), xvid's 2-pass will undersize slightly.

with a custom matrix it's possible that this effect is being exaggerated... so when you denoise the rate-control behaves properly.

...that's my guess, anyway.

[Manao]

How did you configure XviD ?

[Didée]

(Also, unfilter doesn't really count as a denoiser, at best it's just a very mild median cut, for which removegrain(1) is better suited; unfilter is more for a removal of mild oversharpening or blurring.)
Sorry, but that can't be left on its own.
Unfilter() is not at all a median filter. It's a 5-tap blur/sharpen filter, quite similar to Avisynth's internal 3-tap blur()/sharpen() filter. IIRC, Tom altered the kernel a little, trying that unfilter(+/-) shall undo a prior blur/sharpen operation (i.e. unfilter(-100) should make sharpen(1) undone as far as possible).

[Chainmax]

...
What exactly does the info=3 and colormatrix hints do? im not quite clear on that, but does it help them to work together?
...


I don't know what they do, but that's how I read was best to use them. IIRC it had something to do about ColorMatrix using hints from DGDecode (MPEG2Source).

[koolcracker]

When I tried using the info=3 and hints=true, the picture just looked faded, the problem that I was trying to correct with colormatrix in the first place.

[Wilbert]

When I tried using the info=3 and hints=true, the picture just looked faded, the problem that I was trying to correct with colormatrix in the first place.
Please upload a short vob somewhere where this problem is visible. Also describe what coefficients is used in your vob and what your destination format is (XviD i guess).

What exactly does the info=3 and colormatrix hints do? im not quite clear on that, but does it help them to work together?
from the documentation:

hints (default false)

DGDecode v1.20 and newer versions output colorimetry hints in the video. The colorimetry info (see Colorimetry) can be viewed using

Mpeg2source("F:\TestStreams\avs\AguileraGrammies.d2v", info=1)

The hints are used when setting info=3 in Mpeg2source, setting hints = true in ColorMatrix and when using ColorMatrix directly after loading the video

Mpeg2source("F:\TestStreams\avs\AguileraGrammies.d2v", info=3)
ColorMatrix(hints=true)

When hints are not passed through (for example because you are using a wrong dvd2avi/dgdecode version) it will output an error. When using hints=true, the conversion specified by mode will only be applied when the colorimetry of the input video does not match the specified destination colorimetry. For example, if mode is set to "Rec.709->Rec.601", then the conversion will only be applied if the input video's colorimetry is not already "Rec.601". If the input video's colorimetry already matches the destination colorimetry then the frames are passed through untouched (except for limiter). You can see which scenario is happening by using debug=true.

Technically (although I've never seen such streams) the colorimetry info can be different throughout your video, the hints option will handle this correctly.