What would you do with this captured video? (Comb-like problem)

[Asterra]

Here's a sample:



(Be sure to expand the image.) This is what the video looks like after I have done an inverse telecine to generate the 23.976fps result. (This had to be done somewhat manually because the original video changed phase a half dozen times from beginning to end.) There are two issues:

1) The comb-like problem, readily visible in areas of high contrast, such as the edge of the door in the upper right. Alternating scanlines are about two pixels out of horizontal alignment. But the problem recedes and seems to be gone towards the bottom of the image.

2) I have found that reds and browns exhibit a lack of agreement in how bright they're supposed to be, on a scanline basis. Observe the cabinet in the lower right, or the brown floor. One scanline is bright, the next is darker, the next brighter, and so on.

These are particularly noticeable problems because in both cases - the horizontal misalignment and the bright/dark weave - the effect is temporally static, meaning that as the video is being watched, the horizontal problem doesn't ever reverse itself, nor does the bright/dark lines issue (which could have assisted in obscuring the phenomenon). So they are in-your-face and a constant annoyance. Imagine if dot-crawl was static. It's that bad.

I don't suppose there's any real point in trying to determine why this happened. Possibly it is a flaw particular to my VCR, or how the video was recorded to tape. Bottom line is this is what I'm stuck with. So I'm hopeful that there are some esoteric Avisynth plugins that could help. I've been more than a little astonished at my luck in the past, I can tell you. ;p

Edit: There's also an overall graininess typical of heavily-viewed VHS footage. I've had to conclude that there isn't anything that can be done about this without destroying detail better left in.

[poisondeathray]

vinverse for residual combing, but it will soften slightly

[jagabo]

The only way to fix that is to go back and use a S-VHS deck with a line TBC and capture again.

[2Bdecided]

The only way to fix that is to go back and use a S-VHS deck with a line TBC and capture again.

This is absolute correct - the two fields (i.e. each pair of adjacent lines) come off the tape 1/60th or 1/50th of a second apart, and the deck needs to be entirely jitter-free to avoid combing when they're re-combined. VHS often has plenty of jitter, hence the need for a TBC.

Even with a TBC, there's no guarantee it'll be perfect, but it will probably be a lot better.

I don't know of an AVIsynth function that aligns the fields in this instance, but I wonder if using a very good deinterlacer (e.g. QTGMC, or the older mcbob) will work, because it'll motion track the jitter as if it was a moving object. Use selectevery(2,0) afterwards to get back to the original frame rate.

So, IVTC (as you have done), then use QTGMC, then use selectevery(2,0). Unlike vinverse, QTGMC shouldn't soften it at all where it can track the motion. Won't be perfect though, and will be very slow. As you haven't provided a sample, I can't try it for you, so you'll have to experiment.

Cheers,
David.

[manono]

I don't know of an AVIsynth function that aligns the fields in this instance, but I wonder if using a very good deinterlacer (e.g. QTGMC, or the older mcbob) will work, because it'll motion track the jitter as if it was a moving object. Use selectevery(2,0) afterwards to get back to the original frame rate.

So, IVTC (as you have done), then use QTGMC, then use selectevery(2,0). .

I've seen a few like this already turned into DVDs so they can't be recapped. I do pretty much the same as you suggested:

QTGMC( Preset="Medium" )
SelectEven()

But pdr's suggestion of Vinverse is a helluva lot faster and shouldn't soften the areas that don't show interlacing.

There are some AviSynth filters for this, V.C. Mohan's DeJitter and also TBC, but I could never get them to work properly.

[jagabo]

The only way to fix that is to go back and use a S-VHS deck with a line TBC and capture again.

This is absolute correct - the two fields (i.e. each pair of adjacent lines) come off the tape 1/60th or 1/50th of a second apart, and the deck needs to be entirely jitter-free to avoid combing when they're re-combined. VHS often has plenty of jitter, hence the need for a TBC.

Even with a TBC, there's no guarantee it'll be perfect, but it will probably be a lot better.

This type of jitter happens because the VCR can't spin the drum at a constant rate, it's continuously adjusting the speed a little faster and slower to maintain a long term average. So the individual scan lines that come out of the VCR are a little shorter or longer. A better VCR, even one without a TBC, may do a little better than your current VCR. But a line TBC is the best solution.

Some with/without TBC examples:
https://forum.videohelp.com/threads/230650-Confused-Why-a-VCR-with-TBC-if-separate-TBC-...=1#post1358696

One of the images there is from using a DV camcorder in pass-through mode. That camcorder had a built in TBC. There are also a few DVD recorders with built in TBCs that work in pass through mode, the Panasonic ES15, for example:
https://forum.videohelp.com/threads/319420-Who-uses-a-DVD-recorder-as-a-line-TBC-and-wh...=1#post1983288

Be aware there are two types of TBC: a line TBC, and a full frame TBC. A line TBC fixes this horizontal jitter problem. A full frame TBC provides a continuous signal with perfect timing even in the source video drops out. A full frame TBC may or may not include a line TBC.

I don't know of an AVIsynth function that aligns the fields in this instance, but I wonder if using a very good deinterlacer (e.g. QTGMC, or the older mcbob) will work, because it'll motion track the jitter as if it was a moving object.

No, QTGMC won't really work. All the motion tracking filters track motion of macroblocks, not individual scan lines or pixels. It may clean the video up a little by virtue of all the other filtering it does. There has been at least one attempt at a jitter removal script for AviSynth. But it usually screws up the video even more. To really fix this type of problem you need access to the horizontal sync pulse in the analog video signal. Once the video is captured that pulse is gone.

[sanlyn]

Some with/without TBC examples:
https://forum.videohelp.com/threads/230650-Confused-Why-a-VCR-with-TBC-if-separate-TBC-...=1#post1358696

jagabo, thank you. I don't how you find these old posts, but I spent days looking for this one. Thanks again.

[jagabo]

jagabo, thank you. I don't how you find these old posts, but I spent days looking for this one.

I have a collection of links to interesting posts.