DeDot for checkerboarding?

[clashradio]

I have a symptom which I believe is known as checkerboarding? or maybe it's called dot crawl?

It's common with NTSC laserdisc captures. The closest plugin I could find for this is DeDot. I've tried the code for reduce rainbow, reduce dot crawl; nothing seems to help.

I am using a 3D comb filter, which I know can cause artifacts with moving objects as the filter resorts back to 2D. But 3D has more resolution than 2D I've read.

I've uploaded a short clip of a torture test. Check out frame #6

[poisondeathray]

It's a temporal artifact from the next frame

Hardware 3d comb ? Capture one with, one without, and combine the best parts (mix and match)

In this case, you have 1 good field that you can upscale (or deinterlace) and replace the bad frame. But that doesn't mean you can fix all problems in the rest of the capture

Use separatefields() and look for the clean field

(it's ivtc'ed frame 5 that is affected and replaced)

[clashradio]

This particular test was with a dvd recorder as a pass-through for it's 3D comb filter. My player (HLD-X0) also has built-in 3D comb filter, which I think looks nice but everyone in the world says not to use it. I'm able to output Component with the dvd recorder, s-video with the player.

I do not get the checkerboarding effect when using a 2D filter, or no filter at all. But I do get more rainbows with 2D.

Yes, I could go through and manually delete all of the scene-change frames where there is checkerboaring. I was hoping DeDot would help. I have lowered a little of the symptom using Neat Video in post.

[poisondeathray]

But I do get more rainbows with 2D.

rainbows on a single frame of a scene change would look better than this massive defect. And you could probably filter the rainbows semi-effectively. Use the 3D for most frames, replace only the affected frames with the 2D version +/- derainbow.

Yes, I could go through and manually delete all of the scene-change frames where there is checkerboaring. I was hoping DeDot would help. I have lowered a little of the symptom using Neat Video in post.

No filter like dedot or neat video will be very effective. They will cause way more problems like blurring the picture and still leave the artifact. This defect is to large

To be clear, that example above is not deleting the affected frame only (which would cause sync issues). It's replacing it with a single upscaled field . But you might not have that option on other defects (might not have a clean field)

[johnmeyer]

Definitely not dot crawl. Looks like a bad capture. Check your capture device settings.

If your LD player has component, use that. Otherwise, use composite. S-Video has no benefit on a LD player because the video is store in composite format. This is different from the situation with VHS tapes where S-video avoids transferring the video back to composite.

I'd start by capturing without any sort of comb filter or any other device, and make the capture as simple as possible.

[clashradio]

I agree I'd rather have rainbows than this checkerboard issue (or whatever it is). And I'd rather capture the complete project with a 2D filter vs. recap & replaced the bad frame(s).

Capturing on my PC introduces noise that is not there when I view on my tv, that's why I use Neat Video. But I also admit that can cause blurriness and people can look a little plasticky if applied too heavily.

I'm a little confused when you say replacing the bad frame with a single upscaled field? I plan on performing IVTC as this footage was telecinced, and I plan on upscaling to 720P.

johnmeyer: the reason I'd use s-video out from the player is that output has the built-in 3D comb filter. When I use an external comb filter I use the composite out from the player. My cap settings are Uncompressed 8-bit 4:2:2 YUY2, Lanczos, and video buffers at 10 (default is 50). As far as I understand I'm capturing at the best possible settings for max quality.

As stated above I do not get this terrible issue when no comb filter is in the chain, but I thought for composite RF signal you had to have a comb filter.

[johnmeyer]

There were lots of well-meaning circuit designs in old NTSC televisions that were designed to fix issues related to all the engineering tradeoffs, especially those related to how color was shoehorned into the original B&W spec. The comb filter appeared fairly late in the game and was usually found on high-end sets. Some of them were not very good, and as this article describes, "average" comb filter could often make the video look worse:

[edit] Videohelp's profanity filter won't let me post this URL, simply because, if you parse the string, there is a semi-bad word (unless you raise chickens) buried in the middle. Here's a link where you can find out more about comb filters and how they can screw things up (you'll have to remove the spaces and then cut/ paste: thanks Videohelp!! -- it wouldn't even let me bury this string inside some link text.)

http://www.c o c k a m.com/vidcomb.htm#Subtopics

This is analogous to the situation with time base correctors, many of which can degrade, rather than improve the signal.

So, as I said in my previous post, I'd try capturing that same section of video, but without the comb filter or any other "improvement" circuits, and see what you get.

[clashradio]

I admit I don't get any checkerboarding when there is no comb filter in the mix, but check out the attached pix. You will notice nasty rainbows in boxes 300 and 400, and the orb when there is no comb filter. I'm still searching for the ultimate 3D comb filter. My Toshiba dvd recorder gives pretty good results on the test pattern, and has the least amount of checkerboarding so far.

Being how the source is laserdisc, I don't need a TBC in the chain.

I've have stumbled on that page by Allan Jayne Jr., very good info there. I've also read a good review on comb filters at "sapereaumdesa.ga" NTSC Decoding Basics (part 4) Extron