Static rainbow spot on VHS capture - fixable?
 

I just captured a bunch of tapes with a regular VCR and after finishing my captures, I noticed most had a rainbow spot above the overscan towards the bottom of the frame. It appears in the same spot throughout the entire capture, fluctuating colors in a rainbow pattern that is translucent.

I have checked out several rainbow filters but they all seem to apply to random noise, while this issue seems to carry out through the entirety of the tape. I would prefer not to recapture if possible. Is there a fix for this type of rainbow noise/interference?

I am guessing at this point the VCR head was probably dirty. I cleaned it this morning but have yet to test it.

Post a sample clip.
This sounds like dot crawl + chroma noise, which looks like a rainbow. The solution is to use a better VCR.

I have attached a jpeg to map out where the discoloration is taking place. It is generally in this area on most of the tapes. The encoding process from 4:2:2 to 4:2:0 seems to soften it a bit, but it is still noticeable.

It's chroma noise, and it's much worse on the top of that test capture/clip.

Can it be fixed? Yes and no. You can remove the rainbow color using either the "Camcorder Color Denoise" in VirtaulDub, or "cnr" in Avisynth. But the underlying image damage may still exist, just without the color.

It's not really a rainbow either. That's ROYGBIV -- red, orange, yellow, green, blue, indigo, violet.

Chroma noise tends to be red and blue, and the sub-colors pink and purple. Sometimes you get green, but it's actually an inverse of the red. It depends on the VCR playing it back, and is more pronounced on PAL/NTSC conversions.

I've seen this for 20 years now on tapes. I hate it, and it's a pet peeve. There's two solutions:
- better S-VHS VCR with TBC
- DVD recorder using the LSI Logic chipset, as that chipset filters chroma noise

And if all else fails, you can now filter it in software -- something that didn't exist 20 or even 10 years ago.

I tried out CCD with no luck, maybe I'm using it wrong. I tried some other chroma smoothers and I can't really see a difference.

Use the "show noise" option. It's very obvious.

Attachment 3722

I will say, however, the the chroma noise is pretty somewhat harsh in intensity, and it could use a first pass with CNR, or even a stacked CCD pass in VirtualDub. The pink flashing is pretty bad.

You should be masking the overscan anyway...

Attachment 3721

It has very obvious dot crawl as well:

Attachment 3720

There's a decent dot crawl filter that can be used.

The video also needs some noise reduction, both temporally and in-frame (static).

See the attached VirtualDub processing settings file (vcf) and my output (saved as interlaced x264, play with VLC)

This is all eliminated by using a better VCR.

Thank you so much! I've been trying to follow the masking guide you guys have up, I didn't realize I wasn't doing enough. I was afraid to crop off too much, but I get the impression I can probably follow this overscan setting for everything.

I tried the posted VCF on the full version and it looks great, and also tried it out on a separate capture and it worked for that too. I think this does the trick!

Some of the masking methods changed in recent years, because of changes to VirtualDub. You can still use the old method, but the newer one takes less steps.

Old method = Crop with a "null transform", then add a resize to pad it back to the correct size (example: 720x480). Total of 2 filters. The first step is a true crop, but the second step makes it a mask.

New method = Simply use 1 resize filter, and leave the canvas the same size by "letterboxing" it. Then "crop" it so it masks.

You use two-pixel increments for interlaced video. There's also no rule that each side must be cropped the same -- sometimes more should be taken off top, bottom or a side.

Make sense? :)

Yes it does. I was reviewing the letterboxing thread today and wanted to confirm the maximum x-crop for overscan. It looks like 20 on each side but I wanted to double check. Some videos seem to have actual footage in the overscan while others have ugly black squiggles, and I want to make sure I'm only cropping overscan vs. what actually plays.

Correct. About 20 pixels are not seen on TV.