Deinterlacing before editing? Do I double the frame-rate?

[ozymango]

Greetings, all. I've tinkered around with video capturing from time to time over many years, doing it just often enough that I vaguely remember how to set everything up so things sorta work right from the start, but not often enough to get any real true "workflow" down. Basically I just wing it, and if you don't like winging it, you might wanna go on to another post.

That being said, in the past I've mostly just ripped VHS tapes to AVI, then pretty much just burned that AVI (after coverting to mpeg2) to DVD, and that's a formal structure and things either work, or they don't. And I never edited these videos much, just spliced out the commercials or when the tape switches from "Home Movies" to "Kolchak: The Night Stalker" (the "Zombie" episode, I love Kolchak) and anyway to get to the point --

In this modern world, I haven't watched a DVD in months, mostly I just stream movies on our HDTV, or download them to various hard drives, you know the drill. But I'm digging through some old VHS tapes I swear I digitized a long time ago, but can't find where I saved the AVI files, so I'm kinda starting again with some videos. And in doing so, after I end up with a bunch of giant AVI files (HuffYUV codec), my plan is, run these AVIs through Kdenlive and then output the edited files to some mp4 or mkv format or like that.

And in doing this, Kdenlive automatically recognizes these are interlaced videos, so when you output, it automatically deinterlaces them. And there a bunch of deinterlace options, and I suppose I can go through and try every possible deinterlace option and see if they do a decent (enough) job, or not -- and so I'm looking through various online forums on the web (starting with Videohelp!) and searching on "Deinterlacing VHS" and of course there are fourteen million posts, and fifty million comments on the posts, and by the end of any thread the conversation has drifted off into "well frankly I'm not a big fan of the new season of 'Stranger Things,' but still holding out hope for a wow finish" and where was I going with this ... oh, yeah ...

Correct me if I'm wrong, but the general consensus on editing interlaced video these days is -- convert to progressive first, by deinterlacing your AVI with some outside deinterlacer (usually some form of Avisynth script), and then edit your deinterlaced video, and then you don't have to worry about fiddling with any deinterlacing settings later.

Seems reasonable enough -- but if we're talking AVS scripting, man that is a rabbit hold of immense proportions, and let's say I'm a compete tyro to Avisynth, or come to think of it, I'm a tyro when it comes to deinterlacing -- so are there some basic steps where I should begin, and tips and traps on doing any of this?

You'll note my title line: I can't remember where I read this, and I can't find the link, dangit, but there was thread where somebody said, "If you're gonna deinterlace a video that's 29.97 fps, you'll want to double the framerate to 59.94 fps, so that you don't miss any fields, and then when you output you can put that back to 29.97 fps and it'll look good."

That sounds like it makes sense -- but how the heck do I set up any kind of deinterlacer or deinterlacing script (Avisynth?) to do this? Or is this something that it already a "default" somewhere? Is that just a standard known thing, do most deinterlacers just assume that you want output that's 59.94fps from a 29.97fps source?

Anyway I'm continuing to research all this, but with any luck, somebody out there is reading this and going, "Oh, yeah, that's a basic thing and it's only a few lines in a script, here's an example." <--- this would be awesome.

Less awesome would be, "Oh, yeah, that's an easy thing, just search in the forums." Look, I'm quite aware that if I spend hundreds of hours searching the web for answers to all my questions, at some point a mass of information will accumulate in my head until it reaches critical mass, at which point my head will either explode or I will be transformed into a blazing sun that will illuminate the entire world, but in the meantime, please please please don't just say, "This question has been asked and answered a hundred times, just search the forums." That being said, I fully expect that if I get any responses to my post, at least a decent percentage of them will read, "This question has been asked and answered a hundred times, just search the forums."

BUT, if I'm lucky, and anybody out there has a few spare moments, knowledge of deinterlacing options, and at least a mild sense of humor, I figured he or she could help get me started on where to even begin here. If it helps, I'll describe my basic situation:

* I've got several AVI captures of VHS tapes, captured at 29.97fps using the HuffYUV codec, at 704x480. My goal is just to watch these on my HDTV -- so I know I don't need to deinterlace per se for my own needs -- but I also want to maybe share them with people who'll watch them on some computer screen (progressive display). So I'm thinking of running these through my favorite video editor (Kdenlive) and just let it handle the deinterlacing -- unless somebody really recommends that I deinterlace the AVI files first, and then that person says something like, "Here's what I do to deinterlace my VHS captures, before I edit them." Man that would be awesome. But any and all helpful suggestions are appreciated.

Thanks in advance for your time and consideration!

[lollo]

With modern displays, the video needs to be deinterlaced, by the monitor/tv itself or by fedding them with a deinterlaced input. The second provides better quality, with few exceptions of high end TVs performing almost the same (i.e. Sony Bravia Oled A95L and similar).

One of the best way to deinterlace is with a complex AviSynth script, called QTGMC. It has a large set of parameters, which needs to be tuned accordingly to the source.

The original frame rate of an interlaced video is 25 frames per second (PAL) and each frame contains 2 fields (then 50 fields per second); each of them contains an "unique moment in time". When you deinteralace, you interpolate intra-field to convert a field to a frame, then you double the frame rate to 50 (new)frames per second. (Complex scripts like QTGMC feature also inter-field processing, to improve the quality of the outcome, but that's another story).

Post a sample of your capture, and we'll help you with a (basic) AviSynth processing!

[pandy]

Simple editing like cutting (if performed on frame accuracy i.e. each, two fields) or changing signal levels (brightness, contrast, saturation) doesn't require deinterlacing - for editing such as resizing deinterlacing is mandatory.
Deinterlacing done by you has advantage over deinterlacing made by playing device in such that you are in control - you decide what kind of deinterlacing algorithm is used.

[ozymango]

Post a sample of your capture, and we'll help you with a (basic) AviSynth processing!

Thanks so much for this! Attached is a 19 second clip from a VHS tape I shot back in 1987 -- this was videotaped on a portable Panasonic VHS VCR, at SP speed, using a connected "Newvicon" camera, that's a single-tube camera, and I gotta admit the camera wasn't the greatest, the AGC really struggled with any light/dark contrasts, but enough excuses, my only point is -- this capture represents about as good as the actual VHS playback on my TV (using an SVHS VCR), it's a very stable capture, no dropped frames, I used a JVC SVHS VCR for playback, passed the SVideo into a Panasonic DMR-ES25 DVD recorder for nominal TBC stabilization, and captured at 29.97fps at 704x480. Oh, yeah, HuffYUV codec, PCM audio. Let me know if any more details will be useful.

This clip is short but is a pan across a bunch of people, so you can really see all the interlacing jaggies as the camera pans. Hope this is decent footage to help work a script. The full video is about an hour, dunno if you need to know that or not.

Once again, thanks for any and all feedback and advice!

P.S. Oh yeah, if you (or anyone else reading this post) have an tips or advice on any filters or tools or whatever I can run on this video to make it "pop" better, to the eye, I'm glad to hear any ideas!

EDIT: Oh yeah, forgot to mention, this is an NTSC clip. Hope you PAL people can still help me!

[Darryl In Canada]

Hi ozymango. The free software Hybrid by Selur will make your life much easier. It is a front end GUI to do all the deinterlacing and other filter stuff so that you don't have to do all the script stuff. You can watch a beginner's guide here. This covers the basic stuff one can do in Hybrid, of which deinterlacing is the most useful.

[ozymango]

Hi ozymango. The free software Hybrid by Selur will make your life much easier.

Thanks so much for the link, I'm gonna go check this out right now!

[Sharc]

@ozymango: Next time you should better capture to 720x480, YUV 4:2:2. Capturing to RGB is a waste of bandwidth/bitrate/filesize. Also, QTGMC() will require YUV rather than RGB. For now you can convert to YUV in Avisynth like

AssumeTFF()
ConvertToYV16(interlaced=true)

Also, the brights (e.g. the white shirts) are clipped. This may be an issue of the tape (overexposure) though. Keep an eye on the levels/proc-amp settings of the capture device.

[ozymango]

@ozymango: Next time you should better capture to 720x480, YUV 4:2:2. Capturing to RGB is a waste of bandwidth/bitrate/filesize. Also, QTGMC() requires YUV rather than RGB. For now you can convert to YUV in Avisynth like

AssumeTFF()
converttoYV16(interlaced=true)

Also, the brights (e.g. the white shirts) are clipped. This may be an issue of the tape (overexposure) though. Keep an eye on the levels/proc-amp settings of the capture device.

Ummm ... I did capture in YUV 4:2:2 interleaved, with the HuffYUV codec, where are you seeing the RGB information? I just checked the file I uploaded and on my machine it shows as a YUV 4:2:2 video, so if other people can confirm it's showing up as RGB on the upload, I'll try again. But I have no clue how the heck that happened.

Also unless I'm hallucinating, 704x480 is a totally acceptable resolution for 4:3 video, even for DVD production -- I can do 720x480 but why? My goal is 4:3 video and with 720x480 I just end up with more junk on the edges I end up having to crop out. Or do you need to save as 720x480 for the deinterlacing scripts to work most optimally?

Though as for the whites being clipped, I'm gonna say this again and again and again and again:

I am NOT looking for people to tell me that my video looks good, or bad, or whatever. What I AM looking for is, how do you do the best job at deinterlacing this kind of video capture? Now if my video capture has to meet certain visual standards or it can't be deinterlaced, that's cool, but give me the specific numbers I need to match to allow for the best deinterlacing.

For the record, this is a really old tape, c. 1987, and it was filmed by me using a portable Panasonic VCR with a battery pack, the whole thing weighed about 40 lbs., and I had to walk to around it all day while trying to shoot the video with a shoulder-mounted video camera with a "Newvicon" video tube, known for streaking and limited color range, but man, it was a huge upgrade to the ancient 8mm film camera I had. So if you have any issues with the basic quality of my video, or you don't like air shows, feel free to flake off.

[Alwyn]

where are you seeing the RGB information?

MediaInfo shows it:

General
Complete name : H:\Videohelp\Ozymango\Airshow-clip-interlaced.avi
Format : AVI
Format/Info : Audio Video Interleave
Format settings : PcmWaveformat
File size : 252 MiB
Duration : 19 s 887 ms
Overall bit rate : 106 Mb/s
Frame rate : 29.970 FPS
Writing library : VirtualDub2 build 44282/release

Video
ID : 0
Format : HuffYUV
Format version : Version 2
Codec ID : HFYU
Duration : 19 s 887 ms
Bit rate : 103 Mb/s
Width : 704 pixels
Height : 480 pixels
Display aspect ratio : 3:2
Frame rate : 29.970 (30000/1001) FPS
Standard : NTSC
Color space : RGB
Bit depth : 8 bits
Scan type : Interlaced
Bits/(Pixel*Frame) : 10.179
Stream size : 244 MiB (97%)

Don't worry, you're not the first to have this issue with HUFF. I do, it was never resolved, and I no longer use it. Try UTVideo (ULY2).

My goal is 4:3 video and with 720x480 I just end up with more junk on the edges I end up having to crop out.

No, you don't get more junk. You just get a properly-proportioned image.

I am NOT looking for people to tell me that my video looks good, or bad, or whatever. What I AM looking for is, how do you do the best job at deinterlacing this kind of video capture? Now if my video capture has to meet certain visual standards or it can't be deinterlaced, that's cool, but give me the specific numbers I need to match to allow for the best deinterlacing.

Calm down tiger. A simple adjustment for capture will make your videos better. @Sharc is not suggesting you do a 10 year thesis in video capturing. To effectively tell him to eff off with his brightness is pretty silly.

trying to shoot the video with a shoulder-mounted video camera with a "Newvicon" video tube, known for streaking and limited color range, but man, it was a huge upgrade to the ancient 8mm film camera I had.

The 8mm cine film I have digitised looks better than any analogue video I've shot. I too had one of those VHS portapacks and it produced cr@p video, as indicated by your file. Tape is/was more convenient but it certainly wasn't better quality.

Re deinterlacing...

Do I do it? Yes.

When? Straight after capture, before taking the video into my NLE because QTGMC does a better job than my NLE.

Do I double-rate deinterlace? Yes. Why? Gives me more fluid motion (double the frames means less movement between frames).

How do I use QTGMC? Using an AVISynth script after setting it up as per my guide here. I find opening scripts in VDub affords me the most flexibility, ease of output and rough editing before exporting into an AVI for my NLE or MP4 if it's only a simple editing job.

[Sharc]

..... feel free to flake off

Good advice, thanks. Why should I waste my time responding to your unqualified statements