Both IVTC and McBob works. Why?

[VonOben]

Hi, I'm a bit puzzled here. I'm doing a transfer from VHS, and the source is partly interlaced and partly progressive.

I didn't know if the source for this VHS was film or not, so I tried a IVTC and ended up with a great looking 24P, without interlace lines.

Anyhow, I tried to apply my regular McBob-filter chain aswell and received a 50P stream that also looks great. Movement in each frame.

I thought that I would end up with a 50P stream with pairs of identical frames if the source weren't truly interlaced. But if it's truly interlaced a IVTC wouldn't work, right?

I'm missing something somewhere.

How does an IVTC applied to interlaced non-TC material look?

[VonOben]

A 54Mbyte, 16 seconds, DV-sample (http://www.mordvap.net/filer/doom9-sample.avi) of my source, and an alternative download location (http://dl.dropbox.com/u/8582653/doom9-sample.avi).

[Didée]

If the source is "partly interlaced, partly progressive", then MCBob won't create "movement in every frame". On progressive sections, MCBob necessarily (should) output duplicates every other frame.

If MCBob outputs 50fps, then the input is 25fps. If the input is 25fps, then (standard) IVTC won't result in a 24p output, but rather in 20 fps output. (25 /5 *4).


Look for a section with steady motion, ideally a camera pan at moderate speed. Only one method will result in smooth+fluid motion, while the other method will have some sort of stuttering.

If you ultimately can't decide, upload a sample to examine.

Edit -- oh, you already gave up & posted a sample. No bonus points for finding the answer yourself. :-D

[VonOben]

Edit -- oh, you already gave up & posted a sample. No bonus points for finding the answer yourself. :-D

Nah, I haven't given up but I'm looking for second opinions. :) That might not be a very good sample to find stuttering, so I'm letting some other people have a look at the whole source. I can't really see any stuttering that shouldn't be there.

(I thought that IVTC on the wrong material wouldn't remove any interlace lines, and the guide that gave me a 24P might be wrong. Will google it.)

[Didée]

No ... IVTC performed on true interlacing will also remove the combing. But the result will not have fluid motion.

And yes ... that sample is not very well suited to make decisions. The start seems to be a blending of different content (and the "background" content seems somehow strange). The later parts are very hectic, with only 2 or 3 frames from one scenechange to the next.

Anyway, is that Andrew Eldritch (from Sisters of Mercy) that I'm seeing there? This Corrosion!!

[VonOben]

Anyway, is that Andrew Eldritch (from Sisters of Mercy) that I'm seeing there? This Corrosion!!

Yep, it is. It was about time someone did a good transfer of the Wake and Shot Rev 2.0 videos.

My hardware chain (SVHS->TBC->DV) spit out great masters of these two. But the aftermath, the filtering or the avoidance of filtering, the many decisions takes forever. :(

And then I find it hard to get good references. My TV is hooked up to a Boxee Box that does alot of post processing, and therefor doesn't give me the truth. :) The computer screens reacts different than the TV. Etc. Etc.

Wish I hade my old CRT-TV and my old XBMC-system. It was easier to use them as references.

[Didée]

Phew, this thing is irritating. As far as I can see, the video really is 25|50 based. Reducing to 24p film will drop frames that shouldn't be dropped. Visually it's not very obvious, because the video (mostly) is so fast and has so many cuts. But there are a few spots where the missing frames can be noticed indeed.

The video contains both progressive parts and true-interlaced parts. The composing probbly was done at a 50p stage, and re-interlaced afterwards. However the composer didn't care for field cadence. Because of this, the video basically has "dynamic phase shift". And worse: since field cadence wasn't cared for, in several of the blended/faded scenes there are both pase-shifted and NOT-phase-shifted content present in the same frame. That's a problem.

The only way to "keep everything" is to either keep it interlaced as-is, or to bob to 50p. Reducing to 25p or 24 might turn out "visually acceptable", but it will be lossy.

[VonOben]

The only way to "keep everything" is to either keep it interlaced as-is, or to bob to 50p. Reducing to 25p or 24 might turn out "visually acceptable", but it will be lossy.

A very, very big thank you. Then McBob it is! :)

[VonOben]

The next part on this source seems to be pure 24fps film, that has been tranfered to PAL with what seems to be called "2:2:2:2:2:2:2:2:2:2:2:3 (Euro) pulldown" or "24:1 Pulldown" or "12:1 Pulldown".

Is it possible to reverse this back to 24p without losses? I've had no luck with IVTC and telecide when following the PAL-guides. There are duplicateded frames, and the wrong frames are decimated. :/

[zee944]

(Sorry, my bad. I thought you're talking about the full length material.)

[VonOben]

It has already been answered. No, you can't reserve it back to 24p without loss. You can deinterlace the interlaced parts to 25p and eventually slow down the whole thing to 24p, but deinterlacing is lossy by nature.

Forget the standard conversion guides, it's hybrid material.

This question was about a very different source than the sample above. But thank you, now I know this kind of pulldown can't be reversed.

[Didée]

Euro pulldown should be easy to revert.

Haven't used telecide for a longer time, but rather tritical's TIVTC package:


source
TFM() # deactivate if it's frame-pulldown instead of field-pulldown (i.e. when source appears progressive, with one full dup frame in every 25)
TDecimate(25)

should basically do it.

(Arguably one should AssumeFPS(25000,1001) at the start, but hey...)