IVTC of old 8mm capture

[kiwiusa]

All,

Apologies if these seem like "dumb" questions, but I'm struggling to find the most appropriate (I avoided using "best" here) way to IVTC some 8mm film that was converted to PAL VHS.

Q1: How can I tell if the original framerate was 16 or 18 fps?

Q2: Would just using TFM and TDecimate be the right answer?

Here is a sample of the mpg capture.

http://rapidshare.com/files/96873710/8mmclip.m2v

Thanks in advance,

Al

[neuron2]

Q1: How can I tell if the original framerate was 16 or 18 fps? It's pretty crappy. I applied this to field match and clean up combs due to the crappy encoding:

telecide()

Then I stepped through a motion segment and noticed this pattern (2 means the same picture twice in a row, 1 mean once):

22122221222 ...

So that is 11 out of 20 are new pictures. We have:

11/20 * 29.97 => 16.5

So it could be 16fps. That is the method. You could check a longer segment.

Then I did:

fdecimate(16.0)

and it looks reasonably smooth considering the crappy source.

Q2: Would just using TFM and TDecimate be the right answer? It depends on what your target frame rate is. If you need 23.976, then yes. If not, see above.

[kiwiusa]

It's pretty crappy. I applied this to field match and clean up combs due to the crappy encoding:

telecide()

Then I stepped through a motion segment and noticed this pattern (2 means the same picture twice in a row, 1 mean once):

22122221222 ...

So that is 11 out of 20 are new pictures. We have:

11/20 * 29.97 => 16.5

So it could be 16fps. That is the method. You could check a longer segment.

Then I did:

fdecimate(16.0)

and it looks reasonably smooth considering the crappy source.

It depends on what your target frame rate is. If you need 23.976, then yes. If not, see above.

Thanks Neuron,

When you mention crappy encoding, is that related to my capture of the VHS source to mpg or something embedded in the source material?

Al

[neuron2]

It's a bit blocky, suggesting you may have skimped on the data rate. But beyond that the source is just very poor quality.

I shouldn't have said crappy encoding, because it's the source that is the problem here.

[2Bdecided]

That's NTSC, not PAL. If you did start with PAL, you've converted it. Probably best not to, at least not before processing it.

There's a lot of blending in there. Many film frames get spread across four video fields (blended with others on at least two of them).

If you had the original 8mm, you could obtain dramatically better results from it. This doesn't even look like the original VHS.

Cheers,
David.

[kiwiusa]

That's NTSC, not PAL. If you did start with PAL, you've converted it. Probably best not to, at least not before processing it.

There's a lot of blending in there. Many film frames get spread across four video fields (blended with others on at least two of them).

If you had the original 8mm, you could obtain dramatically better results from it. This doesn't even look like the original VHS.

Cheers,
David.

Ooooops - OK now I feel like an idiot. I uploaded the wrong clip. I'll correct it when I get home tonight.

Al

[neuron2]

Ooooops - OK now I feel like an idiot. I uploaded the wrong clip. I'll correct it when I get home tonight. Why did you ask this if you really thought you had uploaded PAL?

Q2: Would just using TFM and TDecimate be the right answer?

I assumed you just made a typo saying it was PAL. Please try to give us the right stuff and explanation. It's no fun wasting time for nothing.

[kiwiusa]

Why did you ask this if you really thought you had uploaded PAL?



I assumed you just made a typo saying it was PAL. Please try to give us the right stuff and explanation. It's no fun wasting time for nothing.

I actually have two VHS tapes (1 PAL and 1 NTSC) and I originally took some small captures of both to compare and worked with the PAL version from then on.

Unfortunately, I uploaded the original NTSC clip by mistake before and I apologize for that.

Heres the PAL version (Much cleaner), but I'm still trying to get the best IVTC of the original 8mm out of it.

http://rapidshare.com/files/97090195/8mmclip_PAL_.m2v

Once again I'm sorry about the confusion.

Al

[neuron2]

The analysis is not materially changed, except that the pattern I sampled was 21212121212121...

assumetff
telecide
fdecimate(16.6)

You ought not to talk about applying IVTC to PAL. I suppose if we had irregular pulldown to map 23.976-> 25, or 24->25, it might be appropriate, but that is rare for commercial products. Usually, IVTC refers to restoring 29.97->23.976.

[kiwiusa]

The analysis is not materially changed, except that the pattern I sampled was 21212121212121...

assumetff
telecide
fdecimate(16.0)

You ought not to talk about applying IVTC to PAL. I suppose if we had irregular pulldown to map 23.976-> 25, or 24->25, it might be appropriate, but that is rare for commercial products. Usually, IVTC refers to restoring 29.97->23.976.

Thanks again for your help neuron2. What would be the correct terminology for what I'm trying to do? ie recover (as close as possible) the original 16fps

Al

[neuron2]

Good question.

Do you know if it was done by repeating frames or repeating fields? For the former, I would just say "decimation". For the latter, I can't really argue against "IVTC". But I live in the NTSC world. What do you PAL people think?

The source is so bad it's hard to tell if there really is field pulldown.

To cover your butt, you could say "restore progressive frames at the original frame rate". :)

EDIT: The more I think about it, the more I think IVTC is fine for this context. Whatever was done was for converting film to television.

[2Bdecided]

It's just a video camera pointed at the image from a film projector. There's no fixed link between the shutter or frame rate of the two, and some video fields consist of a blend of two film frames.

When making such a transfer, if people can set the camera shutter speed manually, they usually pick one that minimises flicker. This usually ensures there will be some blends.

Luckily here, most (all?) frames from the source exist unblended on at least one video field.

When using the same technique for 24fps sources, in PAL land you can find that most film frames are only available blended with others on the output video, making recovery even more difficult.

Cheers,
David.