Is this DVHS to DVCPro clip not actually interlaced?

[ELinder]

Someone gave me a program that he recorded off satellite with his DVHS deck, which he subsequently transferred via he thinks a Canopus box and Final Cut Pro to the DVCPro codec many years ago.

I have no experience with how a DVHS deck works or what satellite transmissions are like. MediaInfo shows it as interlaced bottom field first, but when looking at the main body of the program it looks only telecined. Could a DVHS deck record satellite progressive frames, or does it also interlace the signal?

Erich

[msgohan]

Classic 3:2 pulldown aka telecine is the method of turning film frames into an interlaced TV signal.

MediaInfo can't know when an interlaced stream actually contains "hard pulldown". Only "soft pulldown" using flags could be detected this way.

You can still reverse hard pulldown rather than using a deinterlacer.

Note: I haven't viewed the sample.

[sanlyn]

Thanks for the sample.

Far too much DV multi-stage compression noise to make cleanup workable without destroying most of it. DV threw away 50% of the original chroma, then introduced edge effects by disrupting the original field polarity. There's ugly chroma noise, grainy green discoloration, and noisy gradients that would require very aggressive filtering. That's just too much data loss and compression noise for decent results. These mistakes are some else's fault, but they are reasons why DV and Adobe aren't recommended if one wants restoration on noisy sources. Plus, there's some really bad ghosting, mosquito noise and poor motion rendering that can't be fixed (see image below, which is from a de-telecined progressive frame inverse telecined with Avisynth's TIVTC):

sample frame 20 after inverse telecine:


This distortion might be embedded in the original source, but note that mosquito noise is not analog -- it's strictly a digital compression artifact. As msgohan stated, the clip is telecined with plain 3:2 pulldown. Don't use deinterlace filters on telecined video. Besides some distortion upon interpolating from combined fields you end up with duplicate and triplicate frames, and juddery motion. I wouldn't suggest Adobe for inverse telecine. You can use deinterlacing if you want, but you'll have a field match and decimation project on your hands. Capturing to lossy DV with Adobe even if it was ProRes is bad enough, but adding a further stage of Canopus damage doesn't help. The clip isn't ProRes, it's the inferior Canopus DV codec. Most readers here wouldn't have the codec in their Windows systems, and probably not ProRes either, so they can't work on this sample.

I used this code to open the clip and remove telecine (change the path statement as needed):

vidpath="I:\forum\faq\ELinder\B\"
LSMASHVideoSource(vidpath+"43-HappyH-B_clip.mov")
ConvertToYV12(interlaced=true)
TFM().TDecimate()

You could then use a heavy duty denoiser such as QTGMC with InputType=2, or MCTemporalDenoise with "High" settings, or similar. You'd still need some deblocking and something like GradFun3 to smooth the rough and fuzzy gradient edges that will look worse with final re-encoding. All that filtration will visibly soften the image but hard sharpening might not look at good as you anticipate and could generate posterizing. But this is the kind of trouble and work you can expect from this kind of DV capturing.

After telecine is removed, the video is 23.976fps progressive. If you want DVD or SD BluRay, telecine would have to be reinstated by your encoder and the results encoded as interlaced.

Are you sure the original DAR for this movie is 16:9? I don't think it is. Could be wrong there.

[ELinder]

Thanks for looking at it. I'm familiar with removing telecine with TFM and TDecimate, vs what deinterlacing is for. I just wasn't sure what kind of signal digital satellite sent back around 2000 when he recorded this, nor what going from a DVHS tape thru a Canopus box would do. I'm not sure why you keep mentioning Adobe. He captured it back in about 2006 with Final Cut Pro. I'm using AviSynth/Virtualdub for the frame processing.

Quote:

Originally Posted by sanlyn

Thanks for the sample.
Far too much DV multi-stage compression noise to make cleanup workable without destroying most of it. DV threw away 50% of the original chroma, then introduced edge effects by disrupting the original field polarity.

Do you mean when it went from TFF to BFF in the DV conversion?

Quote:

Originally Posted by sanlyn

to lossy DV with Adobe even if it was ProRes is bad enough, but adding a further stage of Canopus damage doesn't help. The clip isn't ProRes, it's the inferior Canopus DV codec. Most readers here wouldn't have the codec in their Windows systems, and probably not ProRes either, so they can't work on this sample.

DVCPro isn't a Canopus codec, is it? If I remember correctly it was developed by Panasonic and was widely used by news organizations for a while before DVCPro50 came out with 4:2:2 vs the DVCPro 4:1:1.

Virtualdub2 can read the codec as well as FFVideoSource for AviSynth.

Quote:

Originally Posted by sanlyn

After telecine is removed, the video is 23.976fps progressive. If you want DVD or SD BluRay, telecine would have to be reinstated by your encoder and the results encoded as interlaced.

I let DVD Studio Pro do the encode and flag setting for DVDs, it seems to do a good job.

Quote:

Originally Posted by sanlyn

Are you sure the original DAR for this movie is 16:9? I don't think it is. Could be wrong there.

No, it should not be 16x9, he definitely messed that up.

Thanks again,
Erich

[ELinder]

While I have your ear so to speak, there's something I've never been able to figure out. Many plugins like TFM and QTGMC need YV12 or YUY2. Everyone seems to use YV12 (including me) which is 4:2:0, but YUY2 is 4:2:2. Shouldn't we be using YUY2 to stay in 4:2:2 when we're probably capturing with HufYUV in YUY2? Aren't we discarding color information inside AVISynth when we do that?

Erich

[sanlyn]

For some reason I thought Premiere Pro was mentioned instead of FCP. my bad for getting it confused over a couple of days. NLE's are not capture apps, regardless of the "Pro" in the name. They're editors. They make inferior analog source captures, mainly because they aren't designed for it and because their operations are set up for digital media.

The YUY2 version of QTGMC has gone through a great many variations, half of which work only in the designers' machines. One problem with forum restoration posts is that as soon as you mention one of the esoteric versions of YUY2 for QTGMC, the specific support dll's that enable the feature go out of print, can't be found, are changed every few months, and so forth. For YUY2 with QTGMC I use the Avisynth 2.5 version of support files. It would be nice if those esoteric programmers could slow down a bit and design a little backward compatibility or more thorough testing before releasing another version. But apparently some folks just like to spend all their time downloading updates. I often wonder if they ever find time to finish a video project. Meanwhile, my slant on recommending filters is to post versions that do the job and that sit still for a while so newcomers can have an easier time of it. As for whether or not newer versions of some filters are "better", that's not always the case. One case in point is the new TemporalDegrain2 -- faster and removes more grain than the first version. But as many users have demonstrated, Version 2 also clearly removes more detail. Choose your poison.

Avisynth's conversions to and from YV12 are the cleanest in the business, so I generally have no problem with it. But I don't like jockeying back and forth with color spaces. Whenever possible I work in the same matrix before moving on to another when needed. And I don't always use QTGMC or YV12 filters. But the fact is, an awful lot of Avisynth's heavy hitters use Yv12. So when I need it, I follow Avisynth's directives for going there.

Like VHS, DV is past. It was never intended for modification except for cut-and-join edits. Whether it's used as a capture medium or whether it comes as DV source, it suffers when modified even if decoded to lossless media and was never designed for re-encoding. It's my experience that those who slobber praise all over it just don't see that well. Sure, it's cleaner than noisy VHS but it looks plastic, lacks finesse, and comes complete with a set of ready-made artifacts that are difficult to deal with. Fortunately consumer cameras that shoot h.264 nowadays are much better (in the hands of users who know what they're doing). But even h.264 has terrific problems when used for restoration.