Deinterlace bobbing results in flickers, jitters?
 

(For reference, this is sort-of a follow-up from this thread, but it feels more like a separate topic, so I thought I'd start a new thread for clarity and forum organization. Hope that's appropriate.)

My deinterlacing seems to be going fine - it's more of a "bobbing" issue that I'm having.

But... based on my understanding of how deinterlacing and bobbing works, I'm wondering if perhaps what I'm experiencing is just a normal, occasional "side-effect" of the bobbing process, simply due to how line doubling works... if that's the case, I can live with it. But I want to make sure this is all "normal" before I move on.


Here are some samples of the issue I'm experiencing. All are the same clip, same duration - but each at different points in the process.

** The effect that I'm referring to will most noticeable at about the 7-second mark (where the jagged-edged golden Japanese text is scrolling horizontally across the screen) - particularly in the final "bobbed" example.

First: a recording directly off a TV screen, to show what the scrolling text is "supposed to" look like: (maybe unnecessary, but I thought it might be helpful?)
Attachment 18600

Next: the VirtualDub-captured YUY2 AVI file:
(Linked here due to being a bit too large for forum upload.)

Next: the Hybrid-encoded, deinterlaced (non-bob) MP4 version:
Attachment 18601

And finally: an alternate deinterlaced MP4 version with line-doubling / bobbing enabled and input scan overwrite set to "top field first:
Attachment 18602

See how the little tattered/scalloped bits along the edges of the golden Japanese text keep jittering up and down? That's the issue. So again, I wonder - is this just a normal "side effect" of the bob process, due to how it works? Or is there perhaps something different I should be doing with my deinterlacing settings for this particular type of clip? I'm using QTGMC, by the way.

Thanks so much to anyone who made it to the end! :salute:

Your AVI file is top field first, make sure you set the field order correctly when de-interlacing, But the source is weird anyway, It has de-interlacing artifacts from the get go on all those added Japanese special effects, so this is not something you can fix with de-interlacing.

What's the full chain that produced the original AVI file?

It really looks like the capture card didn't save the fields as an "every other line" sort of interlace as I dont' think it's possible for a VCR to output video that way, so really seems like something is happening in the chain. Either that, or the original edited master had those artifacts there to begin with?

Can you share your exact workflow? Device-to-device, software and settings, and post processing settings etc. This looks wrong at the source, like it's captured at the wrong resolution or something.

Thanks so much for all the terrific follow-ups!

** Perhaps I should start with this right off the bat:
If it helps diagnose the problems, here is another quick capture that I just made using the exact same physical setup and exact same PC settings, using a different tape. 17 seconds long.



Interesting... maybe it's because I'm new to this and my eyes aren't properly trained yet, because I just dismissed all of that as "normal" combing artifacts. Also, just to make sure - did you see the first link in my post, showing what this video looks like when played directly from VCR to TV? Because the folks in Japan made some VERY strange design choices on that golden side-scrolling text, and I want to make sure that doesn't throw anyone off who is trying to help me.

So, to be clear that I understand the deinterlacing artifacts you're referring to, and to better "train my eyes" - are you referring to the bits shown in the blue circle I've drawn on this screenshot? Or are you referring to something else that I'm missing?
Attachment 18606


JVC HR-S7900U VCR -> GV-USB2 cable -> Windows 10 Home 64-bit / AMD Ryzen 7 1700 Eight-Core Processor /
NVIDIA GeForce GTX 1060 6GB video card -> VirtualDub 1.9.11


I think I might now see what you mean. :hmm: Again, in trying to train my eyes here and see what you more experienced guys are seeing - referring to the image I posted above with the blue circle... are you thinking the fields weren't properly captured in "every other line" because the "combing" effect (like the example above that I circled in blue) doesn't show as a nice, clean, every-other-line type of "comb?" Just making sure I understand your logic so I can try to absorb some of it. :)

If it helps, here is another quick capture that I just made using the exact same physical setup and exact same PC settings, using a different tape.


JVC HR-S7900U VCR -> GV-USB2 cable -> Windows 10 Home 64-bit / AMD Ryzen 7 1700 Eight-Core Processor /
NVIDIA GeForce GTX 1060 6GB video card -> VirtualDub 1.9.11

I'm not sure which settings might be the culprit, of course, so I took a screenshot of every possible setting I thought you might want to check. Let me know if there's anything else you'd like to see:


My JVC HR-S7900U VCR settings:

Attachment 18607

Attachment 18608




My GV-USB2 cable settings:

Attachment 18609

Attachment 18610




My VirtualDub 1.9.11 settings:

Attachment 18615

Attachment 18616

Attachment 18617

Attachment 18618



Thanks so much, everyone! I truly appreciate the help. :worthy:

The problem with the Japanese scrolling characters is in the source video. Separate the fields and examine you will see characteristic notching artifacts. It has been resampled vertically in a progressive manner, while still in interleaved fields causing the problem. Basically it was resized incorrectly. The artifacts are visible on the screen recording of the CRT as well

If it was a recording or hardware issue on your side, you would expect that to issue affect everything and manifest everywhere - not just the section with japanese characters. So there is no problem on your end. The other video is also unaffected

Ah, so you're saying that it's likely someone on the official production side of this promotional tape made an error somewhere along the way (I'm guessing in transferring that side-scrolling text from computer to VHS), resulting in that ugly notching? That's pretty interesting. Here I just assumed it was an intentional, albeit very strange (and surprisingly unattractive) style choice. But this makes a lot more sense!

It's not like I need this tape to be bobbed in 60fps. So for the sake of preserving this tape the way it was produced (and the way it looked on Japanese video store monitors in 1992), I'll probably just deinterlace it without line doubling and move on. Thanks, man!

This also makes a lot of sense - I always thought it was weird that only this part of the video has a lot of extra blank framing on the sides! So what you're saying answers several questions I've had about that gold text segment of this video. Thanks again!

Yes.
Also "Video Calibration" should always be on, It optimizes the RF signal level based on tape formulation or the condition of the tape due to wear.

Calibration should (almost) always been off, as it makes video worse more than not. It's not like B.E.S.T., which is what you refer to here.

As far as I know the naming differs based on the country, Units made for Asia and Europe called "B.E.S.T", units made for north America called "Video Calibration" which is essentially calibrating the head coils current for best RF signal, Video stabilization in the other hand should be off unless needed.

I will dig deeper in the the technical manuals to find out for sure.

True - from what I can tell, Video Stabilization essentially knocks the video down to 240p to eliminate jitters - like some of the old video games. On a crappy tape played on an old CRT TV, this may not have been very perceptible, but it's pronounced on a modern display.

From the testing I've done, Video Calibration seems to help.