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

[Sharc]

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

BWDIFF should be pretty safe in this respect, no?

[lollo]

A small experiment, comparing a progressive video versus its (unnecessarly) deinterlaced version. The script I used:

video_org=AviSource("C:\Users\giuse\Videos\Acquisizioni\xxx.avi")

not_deinterlaced=video_org.SelectEvery(1,0,0) # repeat frames to match deinterlaced pattern
deinterlaced=video_org.AssumeTFF().QTGMC(lossless=1)

### check difference between original video and original video after QTGMC_lossless ###
difference_not_deinterlaced_deinterlaced=Subtract(not_deinterlaced, deinterlaced).Levels(65, 1, 255-64, 0, 255, coring=false)

stackhorizontal(\
subtitle(not_deinterlaced,"not_deinterlaced",size=20,align=2),\
subtitle(deinterlaced,"deinterlaced",size=20,align=2),\
subtitle(difference_not_deinterlaced_deinterlaced,"difference",size=20,align=2)\
)

Marginal differences: comp_slow.avi

In any case, as opposed to dellsam34 approach, being a purist I would never deinterlace a progressive content (except for fixing defects).
All this was a technical discussion (thanks Bwaak and pdr), not a recommendation at all...

[Sharc]

In any case, as opposed to dellsam34 approach, being a purist I would never deinterlace a progressive content (except for fixing defects).
All this was a technical discussion (thanks Bwaak and pdr), not a recommendation at all...

Thanks for the comparison.
A real life situation would be a DVD with progressive video but flagged as interlaced (typical for PAL movie DVDs). If the player or TV deinterlaces the progressive video respecting the flag one depends on the quality of the applied deinterlacer.

[lollo]

A real life situation would be a DVD with progressive video but flagged as interlaced (typical for PAL movie DVDs). If the player or TV deinterlaces the progressive video respecting the flag one depends on the quality of the applied deinterlacer.

Yes. For the PC player is easy to check, for the TV is a bit more difficult, just watching/comparing videos in real time on a screen is not easy.

[poisondeathray]

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.

Yes - I used the terms "effective resolution" vs. "resolution"

"Effective resolution" is defined here as in terms of lines resolvable detail; not "resolution" as in number of pixels. If you upscale something you increase the "resolution" - as in number of pixels - but generally this does not increase the actual details or measurable lines of detail beyond the original source (exceptions could be some types of machine learning). Deinterlacing (as the term is commonly used) is a form of "upscaling" - where a frame is interpolated from a single field. Thus, you never get full actual vertical effective resolution (as in resolvable lines of detail, and ignoring machine learning) from the progressive source compared to a weave ; the latter which uses 100% of the data instead of 50%.

More, if you apply lossless option in QTGMC there is no difference in the kept “progressive” frame between pre and post deinterlacing operation.

Yes there is a difference, the original fields (scan lines) , although kept, are now offset and displaced. This is the same with yadif which also is a "lossless" deinterlacer - "lossless" defined as keeping the original scan lines.

If you start with a interleaved frame , where [t,b] are the original top field and bottom field, where "i" denotes interpolated, and you apply a "lossless" deinterlacer

[t,b]

for bff becomes

[ti,b] [t,bi]

If you started with a progressive content frame, you now have ~50% of the effective vertical resolution in each of the bobbed frames. If the interpolation is great you might have slightly more than 50%. If there is some postprocessing, you might have slightly less . And this is quantifiable on a resolution chart/video

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

BWDIFF should be pretty safe in this respect, no?

Sometimes; there is a threshold setting which might work ok depending on the content. Also it only possibly weaves current frame . If fields are not aligned - shifted or displaced, you can have a wrong match

BWDIF is a lot closer to what most flat panel TV's do , even older cheap ones - they weave 2:2 and have proper cadence detection for common cadences . That's a big difference between SW deinterlacers. The most common ones - Yadif, bob - those are typical deinterlacers used for software playback - they never weave conditionally

A real life situation would be a DVD with progressive video but flagged as interlaced (typical for PAL movie DVDs). If the player or TV deinterlaces the progressive video respecting the flag one depends on the quality of the applied deinterlacer.

Yes - "25p content in 25i" - and this 99.999999999999999% of the time is played back correctly in PAL areas for DVD, broadcast . The analgous situation is "29.97p in 29.97i" for NTSC areas . If you look at TV review/comparisons sites, one of the cadence pattern tests is 2:2 . Basically 100% of them pass.

[dellsam34]

Yes, I'm aware the fields are not identical, I was just addressing the commercials part, since I don't work with movies just few commercials that are a mix of old telecine scans and video camera production it is not worth the time and effort to split them, process them individually and re-joint them, I should have been more clear, sorry for the confusion.

[lollo]

I should have been more clear, sorry for the confusion.

No problem at all, we know that you take care of your videos; and yes for commercials, I also sometimes full deinterlace everything when there is a small amount of progressive content mixed to majority of interlaced frames

[Skiller]

Something to consider when it comes to deinterlacing: because all interlaced video, including PsF video like movies on PAL VHS and TV, is required to be vertically low-passed to limit interline flicker (aka. Interline Twitter), there is no 100% of vertical resolution to begin with in any case.

[Bwaak]

Something to consider when it comes to deinterlacing: because all interlaced video, including PsF video like movies on PAL VHS and TV, is required to be vertically low-passed to limit interline flicker (aka. Interline Twitter), there is no 100% of vertical resolution to begin with in any case.

Movies on TV likely have 2:2 pulldown (with filtering). Footage from DV and HD camcorders does not have filtering or it is switchable.

[Skiller]

Progressive HD formats do not have it and don't need it due to a completely progressive playback chain (thanks to HDMI and Component). This guarantees there is no way of introducing the flicker. But anything standard definition should be compatible with being sent through a interlaced-only interface and is therefore more likely than not to have the vertical low-pass, even when it's PsF.

I have seen some SD DV camcorders not doing the low-pass in PsF mode, and that becomes a nuisance the moment you play it back via S-Video for example. The very fine vertical detail makes it impossible for any deinterlacer (yes, even QTGMC) to discern detail from motion – and in case of doubt a deinterlacer tends to opt for bobbing to avoid potential combing. In detailed shots, that means basically the entire frame is bobbing and showing severe aliasing.

[Bwaak]

Progressive HD formats do not have it and don't need it due to a completely progressive playback chain (thanks to HDMI and Component).

Movies transmitted over 1080i do not represent a "completely progressive playback chain".

This guarantees there is no way of introducing the flicker.

Are you confusing full-screen flicker with interline twitter?

But anything standard definition should be compatible with being sent through a interlaced-only interface and is therefore more likely than not to have the vertical low-pass, even when it's PsF.

HD is no different. Terminology-wise, I would say that if there is low-pass filtering, it is 2:2 pulldown. If there is no filtering, it is PsF. On another hand, PsF is used in HD world primarily, while in SD world it is simply called "progressive", because there are no 50p/60p SD broadcast formats... well, except for Australia in the early 2000s.

I have seen some SD DV camcorders not doing the low-pass in PsF mode, and that becomes a nuisance the moment you play it back via S-Video for example. The very fine vertical detail makes it impossible for any deinterlacer (yes, even QTGMC) to discern detail from motion – and in case of doubt a deinterlacer tends to opt for bobbing to avoid potential combing. In detailed shots, that means basically the entire frame is bobbing and showing severe aliasing.

Don't play footage from a camcorder on a TV. It is meant for editing. Ingest, edit, render, author.

[Sharc]

This guarantees there is no way of introducing the flicker.

Are you confusing full-screen flicker with interline twitter?

Attached examples of unfiltered PsF, deinterlaced with various deinterlacers.

The TV processes these "25p content in 25i" sources perfectly without twitter (similar BWDIF).

SW deinterlacers applied on same source (attachment Various deint.mp4):
- BWDIF: no twitter, lines are razor sharp
- QTGMC(lossless=1): twitter, flicker
- QTGMC(lossless=2): very minor twitter/flicker
- QTGMC(): no twitter, no flicker, but blurring the horizontal lines
- BOB(): twitter and line thickening (doubling)

For BOB() only see attachment Line Twitter 2.mp4

[Skiller]

Gotta look into Sharcs examples in detail later.

Progressive HD formats do not have it and don't need it due to a completely progressive playback chain (thanks to HDMI and Component).

Movies transmitted over 1080i do not represent a "completely progressive playback chain".

Hence my phrasing of "progressive HD formats". I excluded 1080i.

Are you confusing full-screen flicker with interline twitter?

No, matter of terminology. In post #38 I wrote "interline flicker (aka. Interline Twitter)". I tend to use flicker and twitter synonymously when talking about this particular subject. Screen flicker (refresh rate flicker) is related to the display used, I am aware of that.

But anything standard definition should be compatible with being sent through a interlaced-only interface and is therefore more likely than not to have the vertical low-pass, even when it's PsF.

HD is no different.

Yes it is, except for 1080i. Under which circumstances would 720p and 1080p ever run through a deinterlacer unless when wrapped into 1080i? Please explain.

Terminology-wise, I would say that if there is low-pass filtering, it is 2:2 pulldown. If there is no filtering, it is PsF.

That's your definition.

PsF only means a progressive frame is stored in an otherwise interlaced container. Progressive segmented Frame.

Don't play footage from a camcorder on a TV. It is meant for editing. Ingest, edit, render, author.

Sure, but what happens if I edit it, make a 25p DVD from that and play it on an older DVD-player without HDMI or Component? Same issue! Although I admit it is potentially a benefit of having the choice whether the low-pass is needed so I can add it myself or not. It would not be needed if the video is used for the web or upscaled to 720p/1080p.

[lollo]

Attached examples of unfiltered PsF, deinterlaced with various deinterlacers.

Thanks Sharc, quite interesting!

[Bwaak]

Hence my phrasing of "progressive HD formats". I excluded 1080i.

Yes it is, except for 1080i.

1080i is the only HD format relevant to this thread, which is about transmitting progressive content in an interlaced container. True progressive formats are not relevant.

@Sharc, thanks for testing!

The TV processes these "25p content in 25i" sources perfectly without twitter (similar BWDIF).

You mean, your TV handles "25p content in 25i" sources perfectly without twitter. The problem with TV sets, DVD and BD players is that they are all different. Thorough testing of each and every model is needed, and this is exactly what they did 20-25 years ago, but not anymore. HomeTheaterHiFi has done lots of tests, but after they'd changed their website engine, lots of the old tests are gone. I bought my OPPO DVD player thanks to these tests, but after a recent power outage and then a spike, the player is dead. I need to get a new one, but who will advise me now?

Some info can still be found on Wayback Machine, like 2000 DVD Benchmark. If someone thinks that deinterlacing issues had long been solved, I am afraid they have not.

Regarding BWDIF: it seems the VDub filter is 30p only. Can the Avisynth version do 60p?

[poisondeathray]

- BWDIF: no twitter, lines are razor sharp

Notice BWDif isn't quite perfect - there are some distortions at the top of the circle and square intersect

For BWDif with thin lines, you probably need to increase the threshold . 3 is similar to about what a typical TV would use.

(But "Normal" content for interlaced transmission typically wouldn't have thin lines...and/or would be low passed)

Another option for deinterlacer/weaver is TDeint with tryweave setting . And both have edeint options for when deinterlacing is applied

Regarding BWDIF: it seems the VDub filter is 30p only. Can the Avisynth version do 60p?

Yes avs/vpy versions can

[Sharc]

- BWDIF: no twitter, lines are razor sharp

Notice BWDif isn't quite perfect - there are some distortions at the top of the circle and square intersect

For BWDif with thin lines, you probably need to increase the threshold . 3 is similar to about what a typical TV would use.

Yes. Increasing the threshold eliminated those artifacts. Thanks for pointing this out.

(But "Normal" content for interlaced transmission typically wouldn't have thin lines...and/or would be low passed)

Yes. The artificial thin lines were intentional in this demo to visualize the line twitter effect, simulating small details not exceeding one scanline vertical.

Another option for deinterlacer/weaver is TDeint with tryweave setting

Works perfectly for this example as well. Thanks.

[Sharc]

The TV processes these "25p content in 25i" sources perfectly without twitter (similar BWDIF).

You mean, your TV handles "25p content in 25i" sources perfectly without twitter. The problem with TV sets, DVD and BD players is that they are all different. Thorough testing of each and every model is needed, ....

Agree. I can only speak for myself. Also, most DVD players were pure flag readers outputting just fields to the TV (as opposed to cadence detection found in high end models, outputting progressive frames).
My TV seems to analyze the cadence rather than just obeying the flags. When a pattern changes it requires a number of frames to adapt.

[Bwaak]

My TV seems to analyze the cadence rather than just obeying the flags. When a pattern changes it requires a number of frames to adapt.

I am glad that the Home Theater Hi-Fi test results of DVD players can still be found on Wayback Archive.

The bulk of my knowledge about deinterlacing came from this seminal page: DVD Benchmark: Part 5 - Progressive Scan DVD (October, 2000. Latest Update - March, 2005) .

[Sharc]

The bulk of my knowledge about deinterlacing came from this seminal page: DVD Benchmark: Part 5 - Progressive Scan DVD (October, 2000. Latest Update - March, 2005) .

Yes, this article is a classic.

[Sharc]

FWIW here another example of a line twitter test for various SW deinterlacers applied to a static PsF source (PAL 25p content in 25i).

Edit: same with vertical blur added