Help with VHS video

[kenmasters1976]

I have some VHS tapes I want transfer to AVI. I captured a clip with VirtualDub to test the results and VirtualDub reported 0 frames dropped and 0 frames inserted. Then, checking the clip, I noted that it showed a Telecine pattern, so I tried some simple AviSynth scripts to apply IVTC and get rid of interlaced frames and the ghosting effect they produce, the problem is that by doing so, I get some frames going from this:

http://img30.imageshack.us/img30/5016/source1o.th.png (http://img30.imageshack.us/i/source1o.png/)

to this:

http://img5.imageshack.us/img5/1783/source2p.th.png (http://img5.imageshack.us/i/source2p.png/)

There are some parts of the clip with a random frame looking like this, but others, like the part from the picture, where practically all frames end up like this.

I've searched the forum but couldn't find anything that solves it (or didn't know exactly what to search for). Any suggestions?. Thanks in advance.

P.S.: The IVTC scripts I tried were:


Telecide()
Decimate()


and


TFM().TDecimate()


Both give me the same results.

[Guest]

Post a link to an unprocessed source sample.

[kenmasters1976]

I have a limited upload speed, so I hope this works.

http://www.mediafire.com/?mjdvukmjmdy

Thanks.

[Guest]

How did you capture that? I.e., what capture device did you use?

You have some blended fields. I don't mess with that garbage, maybe someone else can help you.

[thewebchat]

Just by looking at your screenshots, I can tell that the problem is the interlacing detection deciding that "clean" (lol) frames are interlaced and then interpolating them, probably with cubic resizing. The solution is to lower the "mi" parameter in TFM or the equivalent in Telecide.

Edit: After looking at your sample, it doesn't seem like your fields are exactly in phase on the "non-interlaced" frames, so you might be in for a fun time.

[kenmasters1976]

I captured it with an old Conexant Bt878 based card. So the problem is caused by the card itself?.

After looking at your sample, it doesn't seem like your fields are exactly in phase on the "non-interlaced" frames, so you might be in for a fun time. By reading your post and neuron2's, I guess it's just out of my league. Is good to know it, though.

Thanks anyway.

[scharfis_brain]

the footage is b0rked. it has been captured in YV12.
Capture using YUY2 instead.

Also the horizontal sync issues (jitter) makes it tough for the fieldmatcher in static scenes to recognise a valid pulldown pattern.

I suggest to conceal the stairstepping with nnedi2() and hide the mice teeth with vinverse()

this is a quick and dirty script to clean that mess up:

loadplugin("c:\x\tivtc.dll")
loadplugin("c:\x\tivtc.dll")
loadplugin("c:\x\vinverse.dll")
loadplugin("c:\x\nnedi2.dll")
loadplugin("c:\x\removedirts.dll")

avisource("kenmasters1976_sample.avi")
assumetff()
tfm(pp=5, clip2=nnedi2())
vinverse()
temporalsoften(2,15,20, mode=2)
mergechroma(crop(0,2,0,0).addborders(0,0,0,2))
tdecimate()
removedirt()

[kenmasters1976]

Thank you scharfis_brain, I tried your script and it seems to work great. I really appreciate your help.

Thanks again.