AviSynth 2.5x and Convolution3D guide for AVI captures

[jimmalenko]

error code 3, or error Line 3 ?

Virtualdub needs a (I think) VFW codec to open any file. Installing the Panasonic DV Codec should fix the issue.

[Sillyname]

error code 3, or error Line 3 ?

Virtualdub needs a (I think) VFW codec to open any file. Installing the Panasonic DV Codec should fix the issue.

Ohhhh! I have the Panasonic DV codec, I was just weary about installing it. Now I know why I need it.

So now my question is this:

Is there an installer version of the PanaDV codec? The machine I tried to install it on has Windows XP Media Center 2005 on it and does not give the "install"option when I right click on the INF file. Someone told me this may be an admin setting on the computer holding me back but I really don't want to even deal with hunting through administrative settings, as it is not my machine.

[jimmalenko]

Sounds more to me like that the user you're logged on as isn't a local administrator.

[Sillyname]

Sounds more to me like that the user you're logged on as isn't a local administrator.

I installed it via entering a few strings in the registry. Now it will open avi videos but it is still not allowing me to load *.avs files.

I played around with the built in filters but none of them got rid of the white vertical dropouts I have on these old beta tape captures. I need to get Convolution3D working with it.

I got a yuy2 only message so I took out the ConverttoRGB string in the AVS and now I get an access denied error message.

It has to be string related now because the program is responding to individual commands within the avs but not loading all the way. Again, I did get Panasonic DV to work. I can load the same video I am trying to filter through convolution3D, by itself as an avi into virtualdub and even play with some of its own filters. But I cannot get Virtualdubmod to accept the AVS file.

[Sillyname]

Here's my string exactly as it was loaded:

LoadPlugin("Convolution3d.dll")
avisource("D:\test.avi")
SeparateFields()
odd=SelectOdd.Convolution3D (0, 32, 128, 16, 64, 10, 0)
evn=SelectEven.Convolution3D (0, 32, 128, 16, 64, 10, 0)
Interleave(evn,odd)
Weave()
DoubleWeave.SelectOdd()

[FulciLives]

Here's my string exactly as it was loaded:

LoadPlugin("Convolution3d.dll")
avisource("D:\test.avi")
SeparateFields()
odd=SelectOdd.Convolution3D (0, 32, 128, 16, 64, 10, 0)
evn=SelectEven.Convolution3D (0, 32, 128, 16, 64, 10, 0)
Interleave(evn,odd)
Weave()
DoubleWeave.SelectOdd()

So you ar saying that this script does not load into VirtualDubMod?

Can you load the AviSynth AVS script into VirtualDubMod using just this simple script:

LoadPlugin("Convolution3d.dll")
avisource("D:\CAPTURE\capture.avi")

Please note you need to "point" to Convolution3D.dll just as you do the AVI file unless you put it in the SYSTEM or SYSTEM32 folder (I forget which).

If you can't get it to work detail again please exactly what error you are getting.

Also try loading the AviSynth AVS into Media Player Classic and see if that works or if not what error does it give you?

- John "FulciLives" Coleman

[Sillyname]

Here's my string exactly as it was loaded:

LoadPlugin("Convolution3d.dll")
avisource("D:\test.avi")
SeparateFields()
odd=SelectOdd.Convolution3D (0, 32, 128, 16, 64, 10, 0)
evn=SelectEven.Convolution3D (0, 32, 128, 16, 64, 10, 0)
Interleave(evn,odd)
Weave()
DoubleWeave.SelectOdd()

So you ar saying that this script does not load into VirtualDubMod?

Can you load the AviSynth AVS script into VirtualDubMod using just this simple script:

LoadPlugin("Convolution3d.dll")
avisource("D:\CAPTURE\capture.avi")

Please note you need to "point" to Convolution3D.dll just as you do the AVI file unless you put it in the SYSTEM or SYSTEM32 folder (I forget which).

If you can't get it to work detail again please exactly what error you are getting.

Also try loading the AviSynth AVS into Media Player Classic and see if that works or if not what error does it give you?

- John "FulciLives" Coleman

I placed convolution3d.dll in both the system and system32 directories per your instructions.

I'll try dumbing down the string a bit and see what happens.

[Sillyname]

(0, 32, 128, 16, 64, 10, 0)

Are the numbers above the values of the matrix for convolution3D?
Can I enter them into the VirtualDub general convolution filter's matrix?
If so, would they go in left to right, top to bottom?
Just wondering, I'll still try to get AVISynth to pull its weight.

[FulciLives]

(0, 32, 128, 16, 64, 10, 0)

Are the numbers above the values of the matrix for convolution3D?
Can I enter them into the VirtualDub general convolution filter's matrix?
If so, would they go in left to right, top to bottom?
Just wondering, I'll still try to get AVISynth to pull its weight.

I never could figure out the VirtualDub "General Convolution" filter. It has more values to fill in that what Convolution3D uses.

- John "FulciLives" Coleman

[Sillyname]

Nice Virtualdub Filter page...

http://priede.bf.lu.lv/ftp/PUB/Win/MultiVide/VirtualDub/filtri/

[charger2000]

Hi
If I "feed" Media Player Classic with the script it should show the "cleaned" video right ? This could be a nice way to preview the video to restore, trying different settings of Convolution3D.
Please correct me if I'm wrong.

Nik

[jimmalenko]

You're spot on.

[charger2000]

Well, I tried to get the preview with MPC, but it's way too slow..., approx 1 frame every 5/6 sec.. , so I fed CCE with 1 minute of AVI test.

The strange thing is that I can't see any difference in the image beetween the three samples
I've used (0, 6, 10, 6, 8, 2.8, 0) and (0, 32, 128, 16, 64, 10, 0) .

The script is very simple:

LoadPlugin("Convolution3d.dll") 
avisource("D:\myworks\prova.avi") 
SeparateFields() 
odd=SelectOdd.Convolution3D (0, 32, 128, 16, 64, 10, 0) 
evn=SelectEven.Convolution3D (0, 32, 128, 16, 64, 10, 0) 
Interleave(evn,odd) 
Weave() 
#DoubleWeave.SelectOdd()   
ConvertToYUY2(interlaced=true)

If there's any difference, certantly it's not worthing the encode, it take 4 or 5 times the original time !

Don't get me wrong, I'm trying to find out what's the prob here !
I'm sure I'm doing something wrong ...

Nik

[charger2000]

ok...
actually there is difference..

without C3D:


WITH C3D:


but I'm trying to remove those spikes/scratches...

[FulciLives]

I've recently been playing around with a filter called RemoveGrain which you can find at the following website ---> CLICK HERE

There are actually 3 different versions depending on your computer. I use RemoveGrainSSE2.dll because I have an Intel Pentium 4 CPU.

It comes with extensive documentation which is nice compared to other filters I've seen that come with very little.

I've mostly been using RemoveGrain(mode=2) which seems to do some decent cleaning without too much loss of detail. There is also a more aggressive mode which is RemoveGrain(mode=17) but I find it removes a lot of detail BUT I've mostly been using it with PAL DVD sources that I have been converting to NTSC so I am starting with a very sharp source that needs minimal cleaning to begin with so what I am trying to say is that RemoveGrain(mode=17) might be OK with VHS since you need more cleaning than with DVD and have less resolution to begin with anyways so ... it's worth a try.

So far I've only used RemoveGrain with PROGRESSIVE video but I believe (knock on wood) that you would still treat INTERLACED video the same way as you do with Convolution3D

So a sample script might look like this:

LoadPlugin("RemoveGrainSSE2.dll") 
avisource("D:\myworks\prova.avi") 
SeparateFields() 
odd=SelectOdd.RemoveGrain(mode=2) 
evn=SelectEven.RemoveGrain(mode=2) 
Interleave(evn,odd) 
Weave() 
#DoubleWeave.SelectOdd()    
ConvertToYUY2(interlaced=true)

The one nice thing I have noticed is that RemoveGrain(mode=2) is much faster than Convolution3D

- John "FulciLives" Coleman

[charger2000]

thanks FulciLives, I have a long test weekend

Is the value of RemoveGrain linear ? Could I try from to the value of 2 till 17 ?
BTW I'll do my homeworks... 8)

thanks

Nik

[FulciLives]

thanks FulciLives, I have a long test weekend

Is the value of RemoveGrain linear ? Could I try from to the value of 2 till 17 ?
BTW I'll do my homeworks... 8)

thanks

Nik

Nik,

It doesn't work the way you think but if you read the documentation that comes with it you will understand. In my reading of it I got the sense that mode=1, mode=2 and mode=17 were the three most "popular" or "usefull" modes. You can even split it up with something like mode=2,17 to treat the luma and chroma separately. In this example one gets mode 2 applied and the other gets mode 17 applied. I forget which comes first though, luma or chroma.

It's all in the included docs.

- John "FulciLives" Coleman

[charger2000]

HI, I've tried the Removegrain, but I had to insert a line who converts the colorspace:

ConverttoYV12()

that's the result with the value of 17 (BTW I've study the docs)



the encoding is MUCH faster than C3D, but IMHO the rusult is a step behind, look at the black line on the lower part of the helmet, between the "BOSS" and the light blue, is more definite and precise in C3D, the result in Removegrain is a bit "destroyed"

Nik

[FulciLives]

I see only a very subtle difference between the pic using Convolution3D and the pic using RemovieGrain.

I even overlapped the images and tried quickly switching back and forth and still found the difference to be very little.

I think I'm ever-so-slightly leaning towards the Convolution3D image as my favorite of the two images but they are so close to me that if RemoveGrain is that much faster for you then that is what I would use.

In the end I would try to compare a few different screen shots from each and just go with the filter you think looks best but if like me you think they look close enough then go for the one that is faster for encoding.

- John "FulciLives" Coleman

[charger2000]

today I tried several filters: RemoveGrain, Vaguedenoiser, Deen, Edeen, Degrainmedian, Descratch, fft3dfilter
Tried HC Encoder 0.17

I have a screencap for every one...

and the winner is...
Convolution3D with the setting of '0, 32, 128, 16, 64, 10, 0' ....

Actually the real winner is Edeen, is able to remove those damned scratches !! But... the fast moving objects suffer from trails, and doubled shape/outlines...
It's a shame becuase for my videos (races of F1 on VHS tapes) was really good.

C3D will take whole night to encode along with CCE, but that's not a problem for me.

BTW for low bitrates (under 5.000Kbs) HC Encoder is sligthly better than CCE Basic.

Nik

[FulciLives]

today I tried several filters: RemoveGrain, Vaguedenoiser, Deen, Edeen, Degrainmedian, Descratch, fft3dfilter
Tried HC Encoder 0.17

I have a screencap for every one...

and the winner is...
Convolution3D with the setting of '0, 32, 128, 16, 64, 10, 0' ....

Actually the real winner is Edeen, is able to remove those damned scratches !! But... the fast moving objects suffer from trails, and doubled shape/outlines...
It's a shame becuase for my videos (races of F1 on VHS tapes) was really good.

C3D will take whole night to encode along with CCE, but that's not a problem for me.

Why not start a new thread and post your examples ... people are always looking for info about "video noise" filters. I would say put it here but this is really the Convolution3D guide so I think it's best you start a new thread.

- John "FulciLives" Coleman

[charger2000]

Avisynth denoise filters comparison

[wiak]

i get green video
when opening it in vdub

[wiak]

i get green video
when opening it in vdub

anyone

[manono]

Hi-

What's the script you're using?

The usual procedure is to comment out all the filtering and open it in VDub with the bare minimum needed, and then add back the lines one-by-one to see which is causing the problem.

[wiak]

orginal file avi file works fine in vdub
hmm i dont think con3d supports my capture card's output colorspace that is YUY2

LG VHS player > Composite > ATI Theater 550 PRO > iuVCR > uncompressed

am using the one on the first page in the guide
i have the sample problem with diffrent de-noising filters like fft3dfilter, fftgpufilter and so on
orginal file avi file works fine in vdub
hmm i dont think con3d supports my capture card's output colorspace that is YUY2

LG VHS player > Composite > ATI Theater 550 PRO > iuVCR > uncompressed

[wiak]

orginal file avi file works fine in vdub
hmm i dont think con3d supports my capture card's output colorspace that is YUY2

LG VHS player > Composite > ATI Theater 550 PRO > iuVCR > uncompressed

am using the one on the first page in the guide
i have the sample problem with diffrent de-noising filters like fft3dfilter, fftgpufilter and so on
orginal file avi file works fine in vdub
hmm i dont think con3d supports my capture card's output colorspace that is YUY2

LG VHS player > Composite > ATI Theater 550 PRO > iuVCR > uncompressed

0011.jpeg

[FulciLives]

@wiak

Please post the actual AviSynth script you are using.

Open it up in NOTEPAD and dump it here copy and paste style.

There are TWO versions of Convolution3D ... one only works with YUY2 and the other only works with YV12 ... the YUY2 version is the better of the two.

Also please provide details on your source. What format is it and what is the resolution etc.?

- John "FulciLives" Coleman

[wiak]

@wiak

Please post the actual AviSynth script you are using.

Open it up in NOTEPAD and dump it here copy and paste style.

There are TWO versions of Convolution3D ... one only works with YUY2 and the other only works with YV12 ... the YUY2 version is the better of the two.

Also please provide details on your source. What format is it and what is the resolution etc.?

- John "FulciLives" Coleman

looks like is working now
and my source is a 34GB uncompressed avi file in

YCbCr 4:2.2 YUY2, 
720x576, 25.000 fps

[Wile_E]

DoubleWeave.SelectOdd()

Many people use this guide, and I'm afraid that Doubleweave is not needed, and is incorrect for BFF, interlaced video. When I use it with my Canopus ADVC-300 bottom field DV-AVI files, it gives me jumpy video. I either delete it or set it to SelectEven(). What is the purpose of doubleweaving? You already do an Interleave, then a weave() to put the fields back together. Doubleweave() is redundant.

Should this line...Interleave(evn,odd) be changed to Interleave(odd,evn) when I use Doubleweave()?

Here is what I currently use...

SeparateFields()
odd=SelectOdd.Convolution3D (0, 6, 10, 6, 8, 2.8, 0)
evn=SelectEven.Convolution3D (0, 6, 10, 6, 8, 2.8, 0)
Interleave(evn,odd)
Weave()