More deinterlacing issues

[Hypersonic1]

I could still upgrade my capture software. Software like VirtualDub is suppose to be the standard here. However, it was dropping a lot of frames, and it produced really large files. I suppose I could use lossless codecs like HuffyYUV or Lagarith, but they still seemed quite large. It seems like I could run out of disk in an hour with such a high bit rate.

Also, I was wondering if it is worth switching from composite to S-Video. I might have to buy an S-Video cable. My Ezcap has an S-Video port, and so does my Hi8 camcorder and S-VHS VCR (I think).

[johnmeyer]

First of all never capture with a variable framerate. That might have caused some of your initial problems (the freeze frame every 13th frame, for sure).

I want to overlay multiple captures by using multiple video tracks. This would average the pixel values, which would reduce the amount of noise in a logarithmic fashion for each capture.

Yes it will, and it does work.

However ...

Over a dozen years ago, when I first started capturing and restoring media, I too came up with that method to reduce noise. It does work, but the fields and frames keep getting out of alignment because it is almost impossible to get the same capture when you re-capture the same analog source. So, you have to go through the multiple videos and slip one or more of the captures at various points on the timeline to get them to re-align. I am not talking about a few extra minutes of work, but hours of work on a big project, with a very high likelihood that you'll miss a few places where the multiple captures don't align, and then you'll get garbage in your output.

It really isn't worth it, especially since noise reduction software has gotten so good in the past decade.

johnmeyer said: "You cannot losslessly deinterlace a source." manono said: "You can if it's to be followed by reinterlacing it which is what the original question was, and if you save to a lossless codec. You use a bob deinterlacer that keeps one of the fields unchanged (not QTGMC)."

I agree, but if you separatefields or bob, and then re-weave, the end result is not deinterlaced. What the OP was wanting to do was to take 60i and turn it into 30p. That operation always degrades the video. I'm not trying to be argumentative, but I also think the OP may be confused on this point.

And, of course, if you use QTGMC as your deinterlacer, good as it is, if you deinterlace with that and the re-weave, quality will be lost.

My capture setup is creating an MPEG-2, but then I have to use Handbrake to convert to a constant frame rate because variable frame rate + Premiere Pro results in audio sync problems.

As I said above (and I think you are now doing this) you must capture using constant frame rate. Variable frame rate is evil, except for streaming applications. It certainly should not be used for capturing and editing analog video.

A variable framerate MPEG-2 capture? Are you sure about that? Do you have a MediaInfo text file to confirm that? Are you sure you don't mean variable bitrate? You realize, don't you, that creating these intermediate MP4s is just degrading a video quality that wasn't so good to begin with? If you absolutely have to create an intermediate that works with Premiere Pro, can't it be a lossless AVI?

+1 to all of that.

I could still upgrade my capture software. Software like VirtualDub is suppose to be the standard here. However, it was dropping a lot of frames, and it produced really large files. I suppose I could use lossless codecs like HuffyYUV or Lagarith, but they still seemed quite large. It seems like I could run out of disk in an hour with such a high bit rate.

By far the easiest, most consistent way to get great captures from old consumer analog video sources (VHS, Beta, 8mm, S-VHS, Hi-8) is to use the passthrough on a DV or 8mm camcorder. Even better is to use a 1394 capture device that also has TBC built in. Yes, DV doesn't have a great color space, but VHS video has horrible color to begin with, so that 4:1:1 DV colorspace doesn't really matter in this application.

I've captured thousands of hours of analog video via DV and I've never seen one dropped frame. By contrast, when I used to use a capture card, I was having all the same problems you are reporting. I'm not saying it can't work, and I do agree that if you can get it to work, you should be able to get a great capture. However, the DV capture is just so simple, easy, and (close to) foolproof.

Also, I was wondering if it is worth switching from composite to S-Video. I might have to buy an S-Video cable. My Ezcap has an S-Video port, and so does my Hi8 camcorder and S-VHS VCR (I think).

It might help a little. It can reduce dot crawl, and you might get slightly better color, because the hardware doesn't have to try to extract the color carrier from the composite signal. However, this is minor compared to the other issues, and I'd put that off until you have everything else working.

If your Hi8 camcorder has a pass through feature, and if you have a computer that has Firewire (1394) input, I want to once again recommend that you use that for your captures. Everything will suddenly get very, very easy.

[Hypersonic1]

If your Hi8 camcorder has a pass through feature, and if you have a computer that has Firewire (1394) input, I want to once again recommend that you use that for your captures. Everything will suddenly get very, very easy.

This isn't a digital8 camcorder, so it doesn't have it. What kind of camcorder would you recommend that's low cost? My laptop doesn't have firewire. Why can't these things just use USB?

Over a dozen years ago, when I first started capturing and restoring media, I too came up with that method to reduce noise. It does work, but the fields and frames keep getting out of alignment because it is almost impossible to get the same capture when you re-capture the same analog source. So, you have to go through the multiple videos and slip one or more of the captures at various points on the timeline to get them to re-align. I am not talking about a few extra minutes of work, but hours of work on a big project, with a very high likelihood that you'll miss a few places where the multiple captures don't align, and then you'll get garbage in your output.

It really isn't worth it, especially since noise reduction software has gotten so good in the past decade.

I have already tried it. It does work well, but there were a few issues. It is incredibly time consuming working on the timeline with this method. Noise reduction software like Neat Video wouldn't get rid of as much as multiple captures + Neat Video. I am mostly concerned about the VHS-C's; some of them drop a lot of frames. The Hi8's that I have are so much easier because they rarely need to be synced up.

I am going to get the S-Video cable because I think it will help a little.

[2Bdecided]

Noise reduction software like Neat Video wouldn't get rid of as much as multiple captures + Neat Video.

Unless your capture equipment is terrible (which it might well be), most of the noise is on the tape itself. Most of the noise is exactly the same every time you play the tape. Most of the noise won't be removed by averaging multiple captures.

Whereas Neat video (and lots of other noise reduction methods) will happily remove the noise that's on the tape.

FWIW I wouldn't use Neat Video or any other denoiser aggressively. I'd leave most of the noise in there, especially any high frequency luma noise. Low frequency chroma noise can sometimes be removed completely, but VHS looks like plastic when you remove ALL the noise.


Multiple captures are genuinely useful for noise that's different every time (often this is true of comet tails).

OTOH multiple captures rarely align perfectly, and that causes the video to be blurred slightly. This appears to reduce the noise, but you can get this effect with one capture and a simple blur command.


I think you should concentrate on doing normal capturing properly, before considering things that most pros think are too difficult to bother with. I fear you don't yet know how much you don't know.

Cheers,
David.

[johnmeyer]

Once again, +1 to everything David just posted, especially about the caution again using a denoiser too aggressively, and also to the huge practical issues surrounding the averaging of multiple captures.

[Hypersonic1]

Here is a before and after. The before had a single capture and the after had multiple captures. Both of them have the same compression. You would probably notice the difference more if it had less compression, but then they would be much larger files.

[johnmeyer]

Wow, that's a lot of work for almost no reduction in noise. I put both clips above each other on the timeline in my NLE and did an A/B between them while playing. I also looked at selected individual frames. Your capture looks pretty good, but the noise reduction from the multiple captures is most definitely not worth all the effort.

[Hypersonic1]

Again, if I would have shown two different lossless clips, the difference would be slightly more noticeable. However, cleaning the heads of my Hi8 camcorder might also improve results too. Maybe?

There are other ways multiple captures can help. For example, if one capture has missing frames in one part, or an artifact, but does not appear in another capture, one would be patched with another. Neat Video would not be able to correct these type of errors. Of course, this process is really time consuming. I just have a really strong interest to improve the quality of my videos.

Well, this is going off topic. I might have to create a new thread soon. I definitely have some questions about how to avoid quality loss with YUV -> RGB -> YUV and audio not syncing with VirtualDub since I want to ditch my capture card software in favor of lossless capturing.

[Hypersonic1]

I just got off the phone with iMemories. They only have the progressive versions. The people working there seemed quite ignorant. They also won't support more than a 4-hour blu ray, even if the footage is SD.

I am going to have to do this myself. I would have done this in the first place, but I knew almost nothing about video. I have learned a lot so far thanks to these kind of sites.

[2Bdecided]

I can see a slight improvement, especially around the date text. It's averaging out very slight horizontal jitter.

If you keep playing your tapes again and again and again, eventually you will wear them out. And the VCR.

Your captures shouldn't drop frames.

The drop-outs will usually be the same every time you capture, but you will get more and more as you wear the tapes out.

Have you checked the video levels? e.g. histogram(mode="levels")

Cheers,
David.

[Hypersonic1]

I can see a slight improvement, especially around the date text. It's averaging out very slight horizontal jitter.

If you keep playing your tapes again and again and again, eventually you will wear them out. And the VCR.

Your captures shouldn't drop frames.

The drop-outs will usually be the same every time you capture, but you will get more and more as you wear the tapes out.

Have you checked the video levels? e.g. histogram(mode="levels")

Cheers,
David.

That sample was from a Hi8 tape. They are in better shape than the VHS-C tapes.

I haven't gotten to levels yet. I am still at the capturing stage; I am waiting for an S-Video cable to arrive in the mail.

[jagabo]

I could still upgrade my capture software. Software like VirtualDub is suppose to be the standard here. However, it was dropping a lot of frames

This is usually because your hard drive isn't fast enough. Uncompressed SD video capture is too much I/O for older drives, even with newer drives if you're writing to the OS drive. You should capture to a secondary internal drive. External USB2 drives are too slow.

and it produced really large files. I suppose I could use lossless codecs like HuffyYUV or Lagarith, but they still seemed quite large.

Using a fast lossless codec like huffyuv will get the I/O rate down low enough to write to the hard drive without dropping frames. And the codec is fast enough that the CPU doesn't become a bottleneck.

It seems like I could run out of disk in an hour with such a high bit rate.

Lossless SD caps usually run around 30 GB/hr. If you have that little drive space left you need another drive.

[2Bdecided]

I haven't gotten to levels yet. I am still at the capturing stage.

You need to make sure you're not clipping whites or blacks when capturing. You can't fix clipped videos afterwards.


As cruel as this is going to sound, in the long run, the next thing I'm going to type is going to help you if you take it in board...

I have learned a lot so far

In terms of what you need to learn to do this job well, you haven't learned half of it. You're going in the right direction, but you're not there yet. Keep learning.

Cheers,
David.

[Hypersonic1]

In terms of what you need to learn to do this job well, you haven't learned half of it. You're going in the right direction, but you're not there yet. Keep learning.

I didn' t know jack a few months ago. I didn't even know how to use a video editor. Now I am quite comfortable working with Premiere Pro.

How am I suppose to get there? I need a guide that is more comprehensive than the guides on this web site that converts every step from the capture stage to the final production.

I am aware of levels. That sample I sent you was just a capture test. I know that I need to keep the levels touching 0 and 100 IRE, but I have questions about that. I will get there when I am at that stage. I just have so many questions. Every time I learn something new, it results in even more questions. I really appreciate you all helping me. Thanks.