Captured VHS tv recording looks progressive (but should be interlaced)

[dave_van_damme]

It's my understanding that content on VHS is stored as interlaced video. However, I've been capturing a number of movies which I recorded from on VHS TV (completed with ads, bumpers etc) and I'm finding that the movie looks progressive (zero to few jaggies during movement) while the ads look interlaced (noticeable interlacing artifacts).

I've attached a short clip from a movie which was recorded from TV. As you can see, despite there being lots of motion there are no jaggies at all. I know that it's not possible for the movie to be progressive, so what am I seeing here? Is it some form of pull-down?

My goal is to upload the full broadcast of the movie + ad breaks to archive.org (they have a streaming media player on their site). What (if any) deinterlacing/decombing/detelecining should I do? I'll be using either Avidemux, Handbreak or Virtualdub.

Thanks!

edit: capture chain Panasonic VCR > Panasonic dvd recorder quasi-TBC > generic USB RCA capture device > Amarec (Lagarith)

[jagabo]

Movies released as PAL video are often progressive frames even if they are encoded/stored interlaced. That appears to be the case here. Your cap is fine as it is. Just encode progressive.

[Bwaak]

Encoding progressive would ruin the commercials. I would cut out the stupid commercials and encode as progressive. You can upload the commercials separately for those who like this kind of crap.

[dave_van_damme]

Encoding progressive would ruin the commercials. I would cut out the stupid commercials and encode as progressive. You can upload the commercials separately for those who like this kind of crap.

Ideally, I’d like the movie and the crap to be together in one video file as that’s the way it was originally shown on TV.

Would there be any way to encode the video so that the movie remains progressive and the deinterlacer does its job on the adverts?

[lollo]

Would there be any way to encode the video so that the movie remains progressive and the deinterlacer does its job on the adverts?

The "progressive" part is not really progressive; the fields are in the same moment in time (contrary to an interlaced video), but the frame is not self-consistent, because it needs the two fields with even and odd scanlines to be merged.

In any case, if you want to deinterlace the interlaced portion in a video containing both "progressive" (as meant before) and interlaced portions, and do not touch the "progressive" segment, you can use the following:

TDeint(mode=0, full=false, cthresh=9, chroma=false, MI=16, edeint=QTGMC(FPSDivisor=2), emask=TMM(mode=0), slow=2)

The frame rate of the interlaced part is not doubled, so it is not optimal for it, but it allows processing (and playing and encoding) of the two portions simultaneously. Try to play with MI and cthres parameters for the best results.

[Sharc]

Would there be any way to encode the video so that the movie remains progressive and the deinterlacer does its job on the adverts?

Encode it as interlaced. The TV (or other player) will do the deinterlacing of the adverts on the fly.

Concerning the "progressive" part see for example here:
https://en.wikipedia.org/wiki/Progressive_segmented_frame.

[jagabo]

Another way to deal with it is to smart bob *(QTGMC?) the entire video to 50p. You could even detect progressive segments and pass them with simple duplicates, smart bobbing the rest. Something along the lines of:

dups = SelectEvery(1, 0, 0)
bobbed = QTGMC()
ConditionalFilter(dups, bobbed, dups, "IsCombedTIVTC(cthresh=9)")

The benefit of doing it this ways is that exact duplicate frames consume almost no bitrate when encoding with interframe codecs. QTTMC will not produce exact duplicates in the progressive segments.

[lollo]

Yes, in this way you end up with a double frame rate output, with optimal QTGMC deinterlacing for interlaced segments and duplicates of frames for "progressive" segments to match the previous.
I experimented this in the past, but found that at the boundary between "progressive" portions and interlaced portions there are more problems than dealing with single frame rate output. It may be related to my setting or to the nature of motion in the fields.
And being the preference for the movie than the commercial, my choice was the single frame rate output.

But OP can experiment both, and see what is better, or do nothing and simply leave the deinterlacing operations to the player, as Sharc suggested.

[Bwaak]

Ideally, I’d like the movie and the crap to be together in one video file as that’s the way it was originally shown on TV.

Who would be interested in a VHS version of "Home Alone"? Why do you think people will like the ads in the middle of the movie?

Home video recorders were used for two main things: for preserving stuff to re-watch, and for time shifting to skip the pesky commercials. Back in the day, people would meticulously cut ads out of the recorded program, I did that all the time and took special pride with clean cuts, using record-pause mode of my VCR having flying erase head. You are turning the technology upside down, burning the ads in the main program.

I know there are people who like watching old ads, so I would cut them out and deinterlace and put them separately. Everyone is happy, and correct deinterlacing is used for both types of content. If this was a movie I wanted to keep, the first think I would do after downloading it, I would cut ads out.

[Skiller]

@ Bwaak
Not sure if this is the case here, but there are people who are interested in old commercials. People upload their old TV recordings with commercials on purpose to YouTube.

[Bwaak]

there are people who are interested in old commercials

Which is why I suggested to upload them separately. I personally hate all commercials, on TV, on the radio or on the Net.

Anyway, I did not realize that there may be a setting for QTGMC to handle both native interlaced and progressive content in one stream. Usually I convert all mixed content to 50p/60p (depending on region), but I always have felt that progressive content is somewhat degraded this way. So, one time I tried to manually split the captured video and deinterlace it in portions, using either "weave", or IVTC or QTGMC depending on the segment. Then I combined them on a 60 fps timeline and encoded as 60p. It still does not look perfect, because when they edited on video machines for video distribution, they did not care for cadence, so even on 24-fps segments I can see combing at edit points, when the cadence breaks.

Hence, the question: what would be the optimal settings for mixed content (both for 50 and for 60)? Say, I have native 30i, 30p and 24fps content, what the settings would be, which would also take care of the combing at the edit points? Will Jagabo's code above work for 60, or do I need different numbers?

[jagabo]

I experimented this in the past, but found that at the boundary between "progressive" portions and interlaced portions there are more problems than dealing with single frame rate output.

Yes, IsCombedTIVTC will sometimes detect an interlaced frame with very litte motion as progressive. And it will sometimes detect progressive frames as interlaced (especially with VHS's horizontal time base errors). That's why I included the cthresh option -- the threshold for deciding between progressive and interlaced. 9 is the default but the user can specify a higher value (more likely to be detected as progressive) or lower value (more likely to be detected as interlaced). In this case I would probably use a lower value. It's not catastrophic if a progressive frame is detected as interlaced, you'll just get two slightly different frames rather than two identical frames.

[Bwaak]

^ Um, so what would be the best settings for mixed content, say, 30PsF, 24 fps and 30i into 60? Just use QTGMC as is? If I wanted it to internally IVTC 24-fps content and then put it back into 60p stream, but now as 2:3 frames, not fields, AND to weave 30psf portions, AND to bob-deinterlace combing when the cadence changes?

[Skiller]

Way too much effort to get it perfectly perfect, in my opinion. Just use QTGMC with 60 fps output on the entire video and mimic what a TV would do. Remember that probably nobody except you will ever notice, and especially if it's for YouTube for example, there are bigger problems anyways.

[jagabo]

I agree, just QTGMC to 60 fps. This is not some rare movie that's not available any other way. If there are some shots missing from later releases just process those to their native frame rate.

[poisondeathray]

Another option for 23.97p/29.97p/29.97i mixed content into a 59.94p stream would be TDeint using mode=1 and tryweave=true. You can specify QTGMC (or another deinterlacer) for Tdeint's edeint parameter . The benefit of TDeint is it's faster than QTGMC, and more weaves should yield higher quality 23.976p and 29.97p originally progressive content sections than QTGMCing everything (but arguably lower quality true interlaced content 29.97i=>59.94p content sections) . In both cases you can use dedup or similar to decimate and generate VFR timecodes

Why VFR ? Judderless displays are becoming common in consumer space for living room TV and computer monitors . One argument against placing everything in 59.94 is you will still have judder on the hardcoded "23.976p in 59.94p" sections (triplicates, duplicates pattern) , but a VFR unique frame version will not exhibit judder on a judderless display (e.g. 120Hz)

[jagabo]

Oops! Posted in the wrong thread.

[dave_van_damme]

Wow lots of replies! Apologies for taking my time to get back.

Yes, in this way you end up with a double frame rate output, with optimal QTGMC deinterlacing for interlaced segments and duplicates of frames for "progressive" segments to match the previous.

Q1. Would that result in duplicate frames for the movie scenes? If so, is there anything inherently "wrong" with having duplicate identical frames? I ask because I have no idea how to use FFMPEG or AviSynth (one day I'll find time to figure it out) so I'm attempting to process this with AVIdemux or Handbreak. Using Yadif+Bob obviously creates 50fps and deinterlaces everything into unique frames *except* for the movie portions which have duplicated frames. I have installed Hybrid so I can try to deinterlace with QTGMC - which leads to my next question...

Q2. Is it possible to deinterlace the whole video (with QTGMC or Yadif+bob) and have the movie scenes remain untouched (and thus, not duplicated) and deinterlace only the interlaced footage?

[lollo]

Q1. Would that result in duplicate frames for the movie scenes? If so, is there anything inherently "wrong" with having duplicate identical frames?

The answers were already there. To summarize:

option 1) the best in term of quality; uses VFR (variable frame rate) as in post #13; interlaced and progressive segments are processed as such; no penalties; requires dealing with VFR material, not common/easy

option 2) deinterlace interlaced content without doubling its frame rate, as in post #5; progressive segments are not touched, interlaced segments are processed; penalty for interlaced segments because you loose half of the motion.

option 3) deinterlace interlaced content at double frame rate, as in post #7; progressive segments are duplicated in term of frames content, interlaced segments are processed at their best; penalty for progressive segments because you duplicate all the frames.

option 4) do nothing, as in post #6, and leave to Display/Player the task of deinterlacing; worst in term of quality for Display (if you do not own a Sony A90J or similar); worst in term of quality for Player

option 5) deinterlace everything as in post #14; penalty for progressive segment, because you apply useless and degradating processing with QTGMC (which is not "transparent" by default, even with specific set of parameters); worst quality for progressive portion.

so I'm attempting to process this with AVIdemux or Handbreak.

Useless GUIs for this kind of tasks.

I have installed Hybrid so I can try to deinterlace with QTGMC - which leads to my next question...

Good choice to start. Then learn AviSynth.

Is it possible to deinterlace the whole video (with QTGMC or Yadif+bob) and have the movie scenes remain untouched (and thus, not duplicated) and deinterlace only the interlaced footage?

I do not use Hybrid, but I suspect that if you are able to insert a command like in post#5 the anwers is yes.
If you want to use approach as in post #13 it may be more difficult because VFR and more complex processing.
Selur may help with better answers for both techniques.


edit: added option 5)

[dave_van_damme]

option 3) deinterlace interlaced content at double frame rate, as in post #7; progressive segments are duplicated in term of frames content, interlaced segments are processed at their best; penalty for progressive segments because you duplicate all the frames.

What is actually the downside of duplicating progressive frames? Does it have any adverse effect on the appearance of the video when it's played? (e.g. does it the video appear 'slower' or something?). Thanks!

[lollo]

option 3) deinterlace interlaced content at double frame rate, as in post #7; progressive segments are duplicated in term of frames content, interlaced segments are processed at their best; penalty for progressive segments because you duplicate all the frames.

What is actually the downside of duplicating progressive frames? Does it have any adverse effect on the appearance of the video when it's played? (e.g. does it the video appear 'slower' or something?). Thanks!

A deinterlacing procedure "changes" the final frame, even if you start from an already progressive architecture. So is not really a duplication, but a modification. Even using lossless options in QTGMC (where one of the field is not touched and the reconstrcted frame is then the same as if it was not "deinterlaced") there is a "change" because the other field/frame is affected.

How big is this change dependes on how good the deinterlacer is in finding "no motion" across the fields (because the original source in progressive) and to apply then the minimum set of changes.

All this in theory. In practice, the "degradation" introduced by deinterlacing a progressive material at double frame rate can be very minimal, so you have to experiment yourself with your source.

[dellsam34]

What is actually the downside of duplicating progressive frames? Does it have any adverse effect on the appearance of the video when it's played? (e.g. does it the video appear 'slower' or something?). Thanks!

I de-interlace all my video as interlaced, Never bothered with telecine (telecine. telecide ..whatever), More frames never hurt. The same for the pulldown, I leave it alone, more frames means less flicker.

[Bwaak]

I de-interlace all my video as interlaced, Never bothered with telecine (telecine. telecide ..whatever), More frames never hurt.

You lose vertical resolution on progressive content. OTOH, your transfers look better than mine, so who am I to judge.

The same for the pulldown, I leave it alone, more frames means less flicker.

There is no flicker on an LCD screen. Flicker is caused by low refresh rate, not by low image rate. Movie theater projectors flash each frame twice or thrice to reduce flicker.

[lollo]

You lose vertical resolution on progressive content.

Why?

[Bwaak]

You lose vertical resolution on progressive content.

Why?

Because joining two fields without processing produces as good or better results compared to figuring out which hip bone connects to which leg bone. This is clearly noticeable on small details, especially those that are not lines. Take a camcorder, shoot an intricate fabric design with curves and circles in 25PsF mode, then compare the results by deinterlacing to 50p and by treating it as 25p.

The attached picture is just an example of a possible subject

[lollo]

I have no idea of what are you talking about, which is out of contest anyhow.

A filed is 720x288 (PAL). A frame is 720x576. A "progressive" video has the 720x288 fields from the same moment in time and are combined to build a 720x576 frame. An interlaced video is deinterlaced by the player or prior by a software processing to build the 720x576 frame.

In other words, if you take the "progressive" video and force an (unnecessary) deinterlace you do not loose any vertical resolution.

[Bwaak]

I have no idea of what are you talking about

Interesting. I thought you knew what I was talking about.

In other words, if you take the "progressive" video and force an (unnecessary) deinterlace you do not loose any vertical resolution.

If the deinterlacer is smart enough to figure out 2:2 pattern and switch to progressive mode, then you will not lose resolution. If the deinterlacer still tries to deinterlace in "video" mode, then you can lose resolution.

[lollo]

If the deinterlacer still tries to deinterlace in "video" mode, then you can lose resolution.

No. But maybe you mean that the deinterlacer (i.e QTGMC) trying to create the frame may (unnecessarly) blur a little bit the moving shapes, that is what was explained earlier.

[poisondeathray]

In other words, if you take the "progressive" video and force an (unnecessary) deinterlace you do not loose any vertical resolution.

If the deinterlacer is smart enough to figure out 2:2 pattern and switch to progressive mode, then you will not lose resolution. If the deinterlacer still tries to deinterlace in "video" mode, then you can lose resolution.

Yes - if it weaves the matching fields for progressive content - you get full effective resolution; otherwise roughly 1/2 vertical effective resolution .

When you apply "deinterlace" - you only have 1 of 2 original fields in the current frame , and you match it with an interpolated field with yadif, QTGMC, bob etc.... (The "real" matching field of the pair is displaced a frame away when you double rate deinterlace, and that gets paired with it's own "fake" interpolated field) . Deinterlacing is really a form of single field upscaling . On progressive content, deinterlacing basically means discarding 1/2 the information for the current frame and trying to re-interpolate what you just discarded (!)

But for many lower quality, VHS sources, you might not be able to tell the difference - because the effective vertical resolution is low to begin with, lacking fine details in the first place. You can definitely see the difference on higher quality sources with fine details, and measure the difference on resolution charts

[lollo]

On progressive content, deinterlacing basically means discarding 1/2 the information for the current frame and trying to re-interpolate what you just discarded (!)

Master pdr, then the final “resolution” is always the same. Just the “information” inside it is different. More, if you apply lossless option in QTGMC there is no difference in the kept “progressive” frame between pre and post deinterlacing operation.