Repair faults on DV tape

[2Bdecided]

(Note to mods: the similar thread in the DV section attacks the problem from the other side - please do not merge, delete or move - thanks!)

I've searched, but not found anything, so...

I have a few DV tapes that were recorded while the camera tape transport was malfunctioning slightly. They're all recorded in LP mode, which didn't help.

When played back, either via the camcorder's analogue TV outputs, or copied onto the PC via Firewire, sections of the video picture (typically down the right hand side) are often replaced with the same section from the previous frame or so. On still pictures, this is almost invisible, but on pans, or with people walking across the screen, the effect is horrible disjointed blocks.

What I'm seeing looks like the in-camera error concealment kicking in, and it's not great!

So, any advice? For other tapes with small momentary problems, I've either managed to clean the heads or just play the tape again, and get a good capture. However, these specific tapes have errors throughout.

I was wondering if I could catch those error interpolated sections, and use something like mvbob to replace the dumb error concealment with motion interpolated correction?

Cheers,
David.

[Mug Funky]

hmm... those bits should be identical to the same parts on the previous frame. so a mask can be constructed quite easily.

however, interpolating could be a problem... so long as the glitches are 1 frame you'll be okay, but sometimes DV glitches will persist way too long, and cover areas of complex motion, meaning photoshop is the only option :|

you could always put a quality warning at the start :)

[edit]

i posted a framedrop interpolator a while ago. that could be hacked to work with masks instead of conditionals.

[2Bdecided]

I'm getting somewhere over in a thread in the DV forum.

There's a Cedocida mod to say where the errors are, but it doesn't seem to always point to the correct part of the frame with my content. Subtracting one frame from the previous one works quite well, though there are often areas (typically 100% white!) which are the same anyway. Maybe I can use the predicted errors to decide which frames to fix, and the frame difference to decide where in the frame to fix?

Getting a bit out of my depth here - need to give it some thought!

Cheers,
David.