automatically repairing video capture line sync artefacts

[Avi.R]

This capture is from a hi-8 video camcorder. The the video output is a PAL source and was thus captured as interlaced 25fps.
Does anyone know now i can automatically repair this? (see attachment).
The amount, size, and position of the affected lines is dynamic and it's on the source, so I can not 'capture it better'.
It should be clever and still work with black objects at the left hand side.
I am using Virtual Dub so it would be preferable if the solution used that, but not an absolute requirement.

Also, is it correct to do this 'fixing' before the application of any deinterlacing?

[2Bdecided]

You need a hardware TBC (time base corrector) during capture. The AVIsynth filters that attempt to solve the problem after capture are vastly inferior (google AVIsynth TBC).

Alternatives to a stand-alone dedicated TBC include a VCR with TBC built-in, a more tolerant capture device, and looping through certain DVD recorders or camcorders which have a TBC built-in. I would use any one of those in preference to a stand-alone TBC because they usually work better. Plenty of TBC threads here on videohelp.

However, if the image you've posted is a pre-deinterlace snapshot (i.e. a whole frame; two fields weaved together), something very strange is happening, and a hardware TBC may not help you.

Cheers,
David.

[jagabo]

DVD recorder as line TBC:
https://forum.videohelp.com/threads/319420-Who-uses-a-DVD-recorder-as-a-line-TBC-and-what-do-you-use

[Avi.R]

The image posted is direct from the capture (from the camcorders PAL RCA video output) and thus still interlaced. The phenomenon can be seen onboard the camcorders own viewfinder. Its one of those small tapes, not a VCR sized one. The problem is only on particular parts of particular tapes. Thus i conclude it was just recorded that way (for whatever reason)

[jagabo]

Software TBCs only work under very specific circumstances. There must be black borders on both the left and right edges -- that's how they determine where the edge of the active picture is, and therefore how much to shift and stretch/shrink the line. That black border must be darker than the picture content at the edges of the farme. Otherwise they can't tell where the edge of the picture is. In theory the edges are at Y=0 and the picture isn't supposed to dip below Y=16. In practice video captures are all over the place so you can't count on that.

If your video doesn't meet those criteria the output of the software TBC will be worse than the source.

A hardware TBC has access to the horizontal sync pulses in the analog waveform so it can use those to adjust the the line position and length.

[Avi.R]

A TBC does correction by adding even more delay (between the sync and the line), which is the opposite of what i need.

Since it is in the viewfinder and the digital copy, this infers the syncs are already correct (so there is nothing for a TBC to correct). The lines themselves are warped.

[jagabo]

A TBC does correction by adding even more delay (between the sync and the line), which is the opposite of what i need.

No.

Since it is in the viewfinder and the digital copy, this infers the syncs are already correct (so there is nothing for a TBC to correct). The lines themselves are warped.

You would play the analog Hi8 tape to the TBC, then to a digital capture device.

[sanlyn]

A TBC does correction by adding even more delay (between the sync and the line), which is the opposite of what i need.

No.

jagabo is correct. You are thinking in terms of a full-frame TBC, which will not help. Note that the type of tbc recommended is a line-level tbc, which corrects line timing within frames, with little or no effect on frame-by-frame timing in itself. Frame-level tbc's take scan lines as-is, with no correction in that respect. IF you were to use both types of tbc in circuit, the line-tbc should come before the frame-tbc. If the frame tbc comes first, the frame timing will be correct but the line-tbc will see no scan line timing errors within frames, because the preconditioned lines will all appear to arrive at exactly "the same time" -- the line-tbc would therefore do nothing.

[2Bdecided]

You're a bit mistaken about TBCs and should listen to jagabo and sanlyn, but...

Since it is in the viewfinder and the digital copy, this infers the syncs are already correct (so there is nothing for a TBC to correct). The lines themselves are warped.

This is a worry. If the camcorder view finder, a good old fashioned TV fed by the composite output (try it!), and the digital capture all show the fault in the exact same way, then it's unlikely a TBC will fix it. I say this because these devices all handle line syncs in slightly different ways, so you would expect faulty/jittery line syncs to yield a picture that looked different on each device. The two fields lining up perfectly (but both way out from where they should be) on that capture is also a worry - wobbly syncs don't usually do that.

I'd still try a TBC of some sort, just in case. It may even be that the camcorder has a TBC built-in, but it can't cope with this problem (or is creating this problem!) and replacing wobbly syncs with clean syncs, but in slightly the wrong place, hence the picture offset. If so, an external TBC won't be able to fix this - but if you can switch the camcorder's TBC off (never seen this capability myself) it might help. A different playback deck might help.

Cheers,
David.