Viewing interlaced video on a computer monitor - a thought

[colinbr]

A question which seems to crop up quite often is regarding the "feathering" which occurs when interlaced video is played on a progressive computer monitor, and a requirement to play the material so that it looks OK on both computer monitor and TV.

In the PAL world, the field rate is 50 fields-per-sec. Now if we set the refresh rate of the computer monitor to 75Hz then this is exactly 1.5x50.

So .... is it feasible to write a video playback program which could "emulate" an interlaced display by displaying only the upper or lower field during a single screen refresh period of the computer monitor? Such a program would have to receive a trigger/interrupt at the end of each screen refresh and repaint its client area with only one field of the interlaced video.

At 75Hz this would require alternate fields from the video to be repeated, padding the 50Hz up to 75Hz.

Of course, this assumes that such a trigger is available from the underlying API. I haven't written any multimedia programs on MS Windows so it may be that this is level of info is just not available to the program.

What does the team think?

Colin

[Flying dutchman]

AFAIK, this technique is already used in many DVD playing softwares such as WinDVD. It is also known as "Bobbing". The odd and even fields are doubled in height, then displayed alternatingly whereby the lower field is shifted down one line. Ideally one should set the refresh rate to 50 Hz or a multiple of that.

[colinbr]

AFAIK, this technique is already used in many DVD playing softwares such as WinDVD. It is also known as "Bobbing".

Oh well .... perhaps I could interest you in this big round thing that I've just reinvented?

[bitz4brainz]

It is not interlacing per se. The feathering occurs on NTSC video -- with film material -- because of the telecine porocess that's done converting 24fps film to 30fps (60 fields) video. This procedure takes every 4 frames and stretches it to 5 frames/10 fields. The problem is an artifact of how the fields are concocted.

On a progressive-display computer, the player necessarily "deinterlaces" the video, but this by itself does not feally fix the problem. The problem can be fixed by a smart player via "reverse telecining" on the fly. That is what is done with progressive DVD players when feeding progressive scan home theatres. It is also done by good computer-based players.

AFAIK, this effect should not be seen on pure PAL because they simply speed the film up to 25 fps, and encode 2 fields per frame. OTOH, some PAL DVDs are converted from NTSC and really look crappy.

Remember, we are talking about film source. Regular video source shouldn't be affected either way. It can be simply deinterlaced.

Here is a detailed explanation:

http://www.hometheaterhifi.com/volume_7_4/dvd-benchmark-part-5-progressive-10-2000.html

[colinbr]

Thanks for that link. It's very interesting.

I didn't realise what a mess the MPEG flags were in, even for major DVD releases! It makes it even more important to seek out in-depth reviews when I buy my next DVD player.

As I live in the PAL world and view most DVDs on a standard interlaced TV, most of this doesn't matter much to me (yet). There is no 3-2 pulldown judder to worry about.

What I didn't realise is that the flags on PAL disks are often set incorrectly making it difficult to work out which two adjacent fields were from the same film frame. This will be important when when the disk is eventually played on a progressive monitor.

And when you introduce the complications of PAL disks transferred from NTSC source, which may in turn have been converted from film; and somewhere along the line the material has been edited breaking the 3-2-3-2... sequence at edit points; and video clips have been edited together with film clips ..... etc
.... then it becomes clear how difficult it is to resurrect the correct progressive film frames (where the 'clip' was originally progressive/film).

Perhaps the best solution is to extract the MPEG file from the DVD and run it through a RIGOROUS cleanup program which examines each field/frame and decides whether it is progressive or interlaced, and then adjusts the MPEG flags so that they correctly reflect the actual content. As this processing would be done "offline" rather than in real time, the cleanup program would be able to do things like backtrack to exactly where 3-2 cadence discontinuities occur rather than having to re-sync several frames later.

Although it may be necesary to recompress some frames if they are encoded as progressive frames and they need to be separate fields - but I'm not to clear on that.

I quite like the idea of being able to "fix" DVDs like this!

Of course I bet copyright restrictions would prevent you from extracting or "messing" about with the file content, even for the best of motives!

I wonder how much analysing of source material TMPGEnc performs...


Colin

[bitz4brainz]

Originally posted by colinbr
Thanks for that link. It's very interesting.

<snip>

Perhaps the best solution is to extract the MPEG file from the DVD and run it through a RIGOROUS cleanup program which examines each field/frame and decides whether it is progressive or interlaced, and then adjusts the MPEG flags so that they correctly reflect the actual content. As this processing would be done "offline" rather than in real time, the cleanup program would be able to do things like backtrack to exactly where 3-2 cadence discontinuities occur rather than having to re-sync several frames later.

Although it may be necesary to recompress some frames if they are encoded as progressive frames and they need to be separate fields - but I'm not to clear on that.

I quite like the idea of being able to "fix" DVDs like this!

Of course I bet copyright restrictions would prevent you from extracting or "messing" about with the file content, even for the best of motives!

I wonder how much analysing of source material TMPGEnc performs...


Colin

I have been doing analog captures from laserdisc through a DV cam, resulting in clean NTSC video. I use TMPGenc to reverse the pulldown BEFORE encoding, then set the flags to tell the player to re-do the pulldown on playback (when necessary). This leaves little to chance, and results in stunning MPEG even at modest bitrates. You would never realize it wasn't a digital extraction to begin with. My only problem has been to find a DVD authoring package that would accept the progressive MPEGs.

[Technoid]

My only problem has been to find a DVD authoring package that would accept the progressive MPEGs.

Don't seem to be an issue with Uleads products

[colinbr]

I use TMPGenc to reverse the pulldown BEFORE encoding [to MPEG] then set the flags to tell the player to re-do the pulldown on playback (when necessary).

So you're effectively using TMPGEnc as the "cleanup" program I described earlier.

I wonder how rigorous TMPGEnc is at detecting and adapting to progressive sequences in material which may contain lots of discontinuities and "wrong" flags.

Is there a tool which allows you to step through an MPEG file frame-by-frame and display the values of the flags (progressive, top_field, repeat_first_field etc) at each frame/field?

[bitz4brainz]

Originally posted by colinbr
I use TMPGenc to reverse the pulldown BEFORE encoding [to MPEG] then set the flags to tell the player to re-do the pulldown on playback (when necessary).

So you're effectively using TMPGEnc as the "cleanup" program I described earlier.

I wonder how rigorous TMPGEnc is at detecting and adapting to progressive sequences in material which may contain lots of discontinuities and "wrong" flags.

Is there a tool which allows you to step through an MPEG file frame-by-frame and display the values of the flags (progressive, top_field, repeat_first_field etc) at each frame/field?

I would guess that TMPGenc does no "cleanup" per se, because the Sony DV camera is giving it clean material to begin with.

Once, I tried using the same technique on mixed material - video documentary with film clips interspersed. It did no harm, but I didn't see a real benefit either, so I don't plan to do it as a regular practice. I don't know for sure whether TMPGEnc was detecting changes between video and film...maybe it was.

With all the tools, both free and not free, out there, I imagine something exists. Try checking this site:

http://www.dvdrhelp.com/

[bitz4brainz]

Originally posted by Technoid
Don't seem to be an issue with Uleads products

Technoid,

I specifically DID have problems with Ulead's MovieFactory 2 and DVD workshop. They didn't seem to like the 24-frame progressive video. They would produce DVDs that stuttered and froze. In Workshop's case, the timecode was wrong too. If you know what I might have doing wrong, I would like to hear about it, because I would like to use MF2.

[colinbr]

I would guess that TMPGenc does no "cleanup" per se, because the Sony DV camera is giving it clean material to begin with

That's assuming there are no discontinuities in the pulldown cadence on the laserdisc. If TMPGEnc just tested the first few fields and deduced a constant pulldown pattern (or perhaps you can specify a pulldown pattern), then it would not cope with discontinuities occuring later on.

Thanks for the link.