Was there ever a case where deinterlacing would simply not work?

[mazinz]

I am running into a possible issue and before I move on to my final process I wanted to ask it here, since technically if it can be fixed it would be a sort of restoration issue

1. A friend in Norway sent me a vhs tape for a film (official release, PAL format)

2. He also sent me a PAL dvdr made with a PAL dvd recorder deck of this film (though interestingly enough it records as interlaced)

3. The whole point is that I am better at editing, so this foreign release has some scenes not in the US version. My source to mix this with is a US dvd release (that is interlaced and NOT progressive with pulldown)

4. So in the end I was going to convert his vhs tape to ntsc and mix it with the US dvd (to fully re-encode both as one later on)

Here is the problem: The film itself was shot in the US, I noticed some interlacing issues (mainly at a scene change and those being the worst) from the PAL tape. At first I was thinking my multi-vcr was to blame (it was running out of that into the data video-1000 tbc and then into the jvcdrm 100s deck). However when I popped in his PAL made dvdr, the same noise and lines appear on his disc (and in the same spots), which then points to the tape itself.

No deinterlacing method I tried fixed or did anything. I also did try to fix it using a "ghosting" filter, this sort of reduced the lines but did not remove them

The actual PAL vhs tape is not in the best of shape and it seems like the company who made this created a real shoddy ntsc to pal film transfer

If the actual PAL master vhs is screwed because of a crap film transfer would it still be possible to fix this or would it be a case of do the best I can with what I have to work with?

It is far from horrific and technically I can go in and remove the very blatant frames during a scene change for the footage, but curiosity is now getting the best of me and was wondering if there was any fix for that tape?

[jagabo]

Deinterlacing is never perfect so it's entirely possible that some interlaced frames will get through. I suspect you have a poor NTSC/PAL conversion with field blending. That can be fixed (mostly) with SRestore() in AviSynth. It would be best for you upload short segments that show the problems you are encountering.

You can use DgIndex to mark small sections demux them. Mark-in, mark-out, File -> Save Project and Demux Video.

[mazinz]

Deinterlacing is never perfect so it's entirely possible that some interlaced frames will get through. I suspect you have a poor NTSC/PAL conversion with field blending. That can be fixed (mostly) with SRestore() in AviSynth. It would be best for you upload short segments that show the problems you are encountering.

You can use DgIndex to mark small sections demux them. Mark-in, mark-out, File -> Save Project and Demux Video.

I have one short scene in mind that I can up for you to view (it would be in mpeg2 format, using womble to cut it). Would you prefer the PAL dvdr version or my ntsc from the pal tape version?

Also going to send you a PM with the clip link once it is up, it is NOT any offensive material

[jagabo]

It would probably be best if I saw both the NTSC and PAL versions since you're having problems with both.

[mazinz]

I will have a link for you prob in 1-2 hours

[mazinz]

Jagabo

PM sent

[jagabo]

Are you familiar with AviSynth? The PAL video cleans up better than the NTSC video and has better levels. I'd probably start with that. I used the following script:

Mpeg2Source("PAL-vhs.d2v", CPU=2)
AssumeTFF()

Yadif(order=1, mode=1)
SRestore()

BilinearResize(720,480)

That left an occasional duplicate frame, a occasional missing frame, and an occasional blended frame, but was pretty clean overall. (Manono does a lot more of this than me. He might be able to recommend a better script. You should PM him.) That gives you a 23.976 fps video with the same running time as your PAL source so you won't have to adjust your audio. From there you can encode progressive with 3:2 pulldown flags to make an NTSC DVD.

[manono]

Manono does a lot more of this than me. He might be able to recommend a better script.

I don't know that I can do any better, but links to samples and such should be given in the thread rather than privately so others that might be able to help get to see them. This is a public forum. This is directed to mazinz, and not to you, jagabo.

[mazinz]

Manono does a lot more of this than me. He might be able to recommend a better script.

I don't know that I can do any better, but links to samples and such should be given in the thread rather than privately so others that might be able to help get to see them. This is a public forum. This is directed to mazinz, and not to you, jagabo.

I was not sure how the forum here would deal with such (as in what is allowed and what is not) but here you go:

http://www.megaupload.com/?d=TN0KSZ49

PS: Oh I agree with you Jagabo, quite a bit was lost in the conversion. I do have a color corrector that I could have run the tape through first, but in actuality, the ntsc tape is much closer to the color combination used on the dvd. Using the PAL dvdr as a source would be fine, since in the end only 7 mins is getting taken from it (and I would re-adjust the colors to match closer with the ntsc dvd during the final encode)

Thank you for the script. I am not unfamiliar with avisynth, but usually if I do not have to use it I wont. If I am not mistaken I think I can use virtualdub to frameserve the avisynth script (in my case to the older tmpg 2.56)

[manono]

I am not unfamiliar with avisynth, but usually if I do not have to use it I wont.

For this kind of thing you have no choice but to use AviSynth.

If I am not mistaken I think I can use virtualdub to frameserve the avisynth script (in my case to the older tmpg 2.56)

You can open AviSynth scripts directly in TMPGEnc (can't you?), although you'll probably want to do the conversion to RGB in the script.

The basic script jagabo gave can't really be improved upon, unless you want to use a better but slower bobber. I decided you might get one frame done correctly in the PAL sample by masking off the subs for the purposes of the unblending. I found one set of duplicate frames where one should have been dropped but because the subs showed up in the second of the 2 frames, they were marked as being sufficiently different from each other that both frames were kept, at the expense of a unique frame nearby. There are probably other examples of that elsewhere in the complete episode. But, like jagabo, I also found some remaining dupe frames, some unique frames being removed, as well as some residual blending. There's no remaining interlacing, though, which was the subject of this thread.

[jagabo]

Yes, QTGMC() gave better results than Yadif(). But it's much slower.

Mpeg2Source("PAL-vhs.d2v", CPU=6)
AssumeTFF()
QTGMC()
SRestore()
LanczosResize(720,480)

I used LanczosResize() instead of BilinearResize() to get a sharper result. And CPU=6 in Mpeg2Source() to get more deblocking and deringing. Attached is an Xvid AVI of that script and the PAL video (no audio).

TMPGEnc Plus accepts AVS scripts. You could also use HcGUI -- which is a better encoder.

[mazinz]

Thank you both for the information. I like the result with the Lanczosresize. It does look much better than before. Thank god only 7 mins is needed from that version of the film. I do have hcGui (as it comes with the FAVC package which I sometimes use depending on what project it is). Now I can properly get back to work on this, thanks again--