deinterlace problem ?

[Byungsoo]

Hi, all

I have a DV source(14G) edited in vegas video and
I'd like to make a divx using that DV.

At the beginning, I used only Virtualdub and internal filters(resize-640x480,smartdeinterlace-default value).
That was very good except colors.
The color was different from my edited DV and I cannot accept it.

Then, I tried Avisynth and filters as followings.

LoadPlugin("c:\program files\avisynth2\plugins\Decomb.dll")
LoadPlugin("c:\program files\avisynth2\plugins\Msharpen.dll")
LoadPlugin("c:\program files\avisynth2\plugins\TomsMoComp.dll")
DirectShowSource("TEST2.avi")
#TomsMoComp(0,15,1)
FieldDeinterlace(blend=false)
BicubicResize(640,480)

This script fed to VD and fast recompressed.

There was the good and the bad.

The good is avisynth make better colors than vd did.
It's similar to DV source.
But , the bad was a little more "blur" or less deinterlaced, I think.

FD was better than TMC but worse than VD smartdeinterace.

In the above 2 cases(VD and avisynth), I used divx5.02.

"Pre processing" option Extreme , 1 pass quality based 100%.

I guess the reason to this results is the followings.

1. When using vd smartdeinterlace, divx5's denoised result fed to
that.( divx5's preprocessed result -> vd's SD. is it right? ).
So,deinterlaced output was clean.
In avisynth case, deinterlace was performed before denoising.
or
2. Tomsmocomp or fielddeinterlace is not as good as vd smartdeinterlace.
or
3. I need to change the parameters for Tomsmocomp or FD.
but, How ?

I am a beginner and my guess is not likely to be right.
But that's my limitation.

Any help will be appreciated.

PS.
I'd like to find a solution using avisynth because it's faster and
make color unchanged.
And I need a conv3d parameter setting for DV,please.

[bb]

Originally posted by Byungsoo
At the beginning, I used only Virtualdub and internal filters(resize-640x480,smartdeinterlace-default value).
Did you read "Q3: How do I properly resize a DV video?
" in the sticky thread "*** F.A.Q. ***"

LoadPlugin("c:\program files\avisynth2\plugins\Decomb.dll")
LoadPlugin("c:\program files\avisynth2\plugins\Msharpen.dll")
LoadPlugin("c:\program files\avisynth2\plugins\TomsMoComp.dll")
DirectShowSource("TEST2.avi")
#TomsMoComp(0,15,1)
FieldDeinterlace(blend=false)
BicubicResize(640,480)

I usually use simply FieldDeinterlace().

"Pre processing" option Extreme , 1 pass quality based 100%.

You critisize the blurring of the deinterlacers, but at the same time you set "Pre processing" to extreme? :confused:
You should use a max. setting of "light".

I guess the reason to this results is the followings.

1. When using vd smartdeinterlace, divx5's denoised result fed to
that.( divx5's preprocessed result -> vd's SD. is it right? ).
So,deinterlaced output was clean.
In avisynth case, deinterlace was performed before denoising.
or
2. Tomsmocomp or fielddeinterlace is not as good as vd smartdeinterlace.
or
3. I need to change the parameters for Tomsmocomp or FD.
but, How ?

1. Wrong. The filter processing occurs before the frames are handed over to the codec. DivX's pre-processing is applied afterwards. This is the case for both, VirtualDub filters and AviSynth.
2. I'd say that's a matter of taste. Any opinions?
3. The defaults should work, and the rest is tweaking and probably depends on the video content.


This is what I can tell, so far. There were some Convolution3D setting recommendations on some threads; please use the search function. The current version has some named, predefined settings. Please refer to its readme.

I hope that helps.

bb

[Byungsoo]

Thankyou, bb

- I have misunderstood vd filter and codec process.

- I read sticky before. But when trying 640x464, bad quality returns.
However, I will try again at home later.

- I have searched for several days but I could not find the conv3d setting for DV.

Anyway, I will test some more case and report the results to this thread.

Thankyou.

[bb]

You could try the Convolution3D "movieHQ" or "movieLQ" settings:
(from the readme)
Convolution3d (preset="movieHQ") // Movie Hi Quality (good DVD source)
is an alias for Convolution3D (0, 3, 4, 3, 4, 2.8, 0)
Convolution3d (preset="movieLQ") // Movie Low Quality (noisy DVD source)
is an alias for Convolution3D (0, 6, 10, 6, 8, 2.8, 0)

- I read sticky before. But when trying 640x464, bad quality returns.
:confused: :confused: :confused:
Why do you get "bad quality" at 640x464, but "good quality" at 640x480? Which resizing filter do you use?
If 640x464 is not an option for you, try to crop 16 pixels from each side and resize to 640x480.

bb

[theReal]

If you know good settings for Smartdeinterlace then you can use the avisynth version of smartdeinterlace. It takes similar arguments than the VD version and it's quite good.

However, for best compression of DV material I'd use Fielddeinterlace() - and btw. there are a lot more options for Fielddeinterlace() that make it behave differently from the default settings.

Then, I'd rather use low pre-processing settings and quant 3 or 4 for encoding - you'll end up with better quality and probably less file size.

[Byungsoo]

Thankyou the Real.

Originally posted by theReal
If you know good settings for Smartdeinterlace then you can use the avisynth version of smartdeinterlace. It takes similar arguments than the VD version and it's quite good

However, for best compression of DV material I'd use Fielddeinterlace() - and btw. there are a lot more options for Fielddeinterlace() that make it behave differently from the default settings.

-> I have tried Tomsmocomp(0,15,1) and it looks good. But I will try your advice.

Originally posted by theReal
Then, I'd rather use low pre-processing settings and quant 3 or 4 for encoding - you'll end up with better quality and probably less file size. [/B]

--> I have used Convolution3d(preset="movieLQ") and unchecked preprocessing. Is it ok if I use the 2 (conv3d and pre processing) at the same time ? I wonder about severe blurring.

And..
do you mean 1-pass quality based encoding ? What do you think is the difference of quality between 3.5 (95%) 1 pass quality and 2 pass 4000 kb encoding ? I don't know what's better under the same resulting size assumption.

Thank you and sorry for my broken English.