Facts about interlace/progressive videos

[chrishallowell]

Hi all,
I spent hours reading through documentation and the messages here about interlacing.
I am trying to sum them up here
Can someone confirm whether or not the below statements are true? Feel free to add any other facts...

1. If you want video to playback on a PC monitor, then the video needs to be progressive.
2. If you want video to playback on a TV, then the video needs to be interlaced.
3. You loose video quality if you convert from interlace to progressive. Smart bob functions will convert video from interlace to progressive.
4. You can use spatial and temporal filters on interlace without quality loss as long as you don't convert it to progressive. Use SeparateFields(), Interleave(), and weave() functions. You can use the below script without degrading video:
assumeTFF.separatefields()
a=selectodd.<<Insert a filter here>>
b=selecteven.<<Insert a filter here>>
interleave(b,a).Weave()

Thanks,
Chris

[manono]

1. No. If it's interlaced, all the players with which I'm familiar can deinterlace it and make it progressive, with varying degrees of quality.
2. No. An interlaced TV set will play the fields and a progressive TV set will play progressive frames.
3. Yes, although that's a tricky one. Deinterlacing creates artifacts that lessen the quality, that's true. It also effectively cuts the framerate in half and makes the video play less smoothly (unless some kind of a bobber is used to deinterlace). On the other hand, interlacing is much harder to compress than is progressive video, so if space is at a premium, it might make sense sometimes to make it progressive. Also, most XviDs and DivXs are made progressive before encoding, if necessary. Bobbing is just one kind of deinterlacing, but, yes, it'll make it progressive. Also, I'm assuming you know the difference between a deinterlace and an IVTC. There's no quality loss when the material is made progressive by way of an IVTC, if it can be IVTC'd.
4. No. Filtering of the kind you're referring to almost by definition creates quality loss. Or at least a quality change. But I don't filter interlaced material that way. Better is to Smart Bob, filter, and then reinterlace. It's better than just separating the fields. One example would be this:

AssumeTFF()
LeakKernelBob(Order=1)
#Filter here#
SeparateFields()
SelectEvery(4,0,3)
Weave()

Other people may have other things, perhaps even contradictory things, to say in response to your questions.

[jagabo]

1. If you want video to playback on a PC monitor, then the video needs to be progressive.

No. You can use a player that deinterlaces on the fly. If you don't have control over the player (eg, you're providing video to others) you're better off deinterlacing.

2. If you want video to playback on a TV, then the video needs to be interlaced.

No, progressive video can be played on a TV. A DVD player will send one field at a time to an SD TV.

3. You loose video quality if you convert from interlace to progressive.

Yes.

Smart bob functions will convert video from interlace to progressive.

Yes. Quality varies.

4. You can use spatial and temporal filters on interlace without quality loss as long as you don't convert it to progressive. Use SeparateFields(), Interleave(), and weave() functions. You can use the below script without degrading video:
assumeTFF.separatefields()
a=selectodd.<<Insert a filter here>>
b=selecteven.<<Insert a filter here>>
interleave(b,a).Weave()

The trouble with this type of filtering is you end up treating scanlines that are not spacially adjacent as if they are. You also are treating fields that are not temporally adjacent as if they are. That's not so say the results can't be acceptable.

[chrishallowell]

Ok, I failed... 50%
Thanks guys for correcting me. I thought DVD format must be interlaced.

Since my source video is interlaced:
It then becomes a question of whether or not it's worth the quality loss to convert it to progressive to use all of the neat avisynth scripts out there. (Most scripts only work with progressive filters.)
I guess that comes down to trial and error.

Thanks,
Chris

[edDV]

Ok, I failed... 50%
Thanks guys for correcting me. I thought DVD format must be interlaced.

Since my source video is interlaced:
It then becomes a question of whether or not it's worth the quality loss to convert it to progressive to use all of the neat avisynth scripts out there. (Most scripts only work with progressive filters.)
I guess that comes down to trial and error.

Thanks,
Chris

Most broadcast video source is interlaced (e.g. NTSC, PAL, 480i, 576i, 1080i) but much of that began with 24fps progressive film as a source. I'll skip over PAL since you are USA. Progressive film is converted to NTSC or 480i/1080i component through a process called telecine and in many cases can be inverse telecined (IVTC) back to 23.976 fps progressive frames. DVD allows either 480i at 29.97 fps or 480p at 23.976 fps (see What is DVD above left).

This document explains telecine and roughly describes how progressive DVD players work (and by extension digital progressive TV).
http://www.dvdfile.com/news/special_report/production_a_z/3_2_pulldown.htm

That document was written around 2001 when progressive players were being introduced. Today inverse telecine or deinterlace is done in hardware chip sets in advanced computer display cards, progressive DVD players and progressive HDTV sets. There is no reason to deinterlace in software if you just want to record to DVD MPeg2 and display. Recording 480i or 1080i is less destructive. Modern hardware deinterlacers are much better than most deinterlace software because they use sophisticated IVTC and motion adaptive deinterlace techniques that are either unavailable (patents) or would take far too long in software.

There are times when software deinterlace is needed. Some justifications are

1. Software IVTC to 23.976 fps will allow 20% more recording time to a DVD at the same quality by removing repeated fields.

2. Many filters or translation effects require deinterlace as a first step.

3. Many MPeg4 (or wmv, flash) compressors must deinterlace to achieve extreme compression. There is a lot of work going on with VC-1 and H.264 to get acceptable interlace compression. 480i and 1080i broadcast will be with us for many years.

4. Some amateurs deinterlace to make the video look better on a computer display. This isn't needed if you use a deinterlacing software player like VLC, WinDVD or PowerDVD. Most premium display cards also deinterlace in hardware when properly set up.

[manono]

I thought DVD format must be interlaced.

As edDV patiently explained, the DVD player output is interlaced. But you can encode for progressive.

Most scripts only work with progressive filters.

Or, most filters only work on progressive video. Yes, that's true, but it's easy enough to take an interlaced video source, bob it to make it progressive at double the framerate, filter it, and then reinterlace it for encoding. I gave an example earlier.

Doing it that way you have to be very careful of the field order - the starting field order, the field order coming out of the script, and the field order set up in the encoder. If it plays jerky coming out of the encoder, you messed up somewhere along the way. One way to check is to add in an AssumeTFF().SeparateFields() as the last line of the script (temporarily). Test the script in VDub(Mod). If it plays smoothly, you have a TFF script and TFF should be used in the encoder. If it plays very jerkily, you finished with BFF, which also has to be taken into account.

[edDV]

"NTSC" DVD can be encoded 480i @ 29.97fps or 480p @ 23.976

DVD recorders typically only record 480i but can play either type of disc.

Standard DVD players play 480i to 480i 29.97 fps (composite, S-Video or YPbPr).

Standard DVD players telecine 480p encoded DVDs to 480i 29.97.

Progressive DVD players can also deinterlace 480i to 480p 59.94 fps over YPbPr or HDMI.

Progressive DVD players can also play 480p 23.976 fps DVD to 480p 59.94 over YPbPr or HDMI using a 323232 frame repeat pattern. This is known as progressive mode. Progressive DVD players do not play at 23.976 fps*.


* this is the hot new feature for Blu-Ray HD players and the new Sony XBR5 HDTV sets. The player sends raw 23.976 (25 pal) fps and the XBR6 HDTV frame repeats 5x to 120 ( 119.88 ) fps. This is supposed to smooth motion judder vs. 323232 59.94 fps and reduce flicker.

[jagabo]

* this is the hot new feature for Blu-Ray HD players and the new Sony XBR5 HDTV sets. The player sends raw 23.976 (25 pal) fps and the XBR6 HDTV frame repeats 5x to 120 ( 119.88 ) fps. This is supposed to smooth motion judder vs. 323232 59.94 fps and reduce flicker.

The new 71 and 81 series Samsung's have this too. Not only can they repeat every frame 5 times, all of them can also use motion vector analysis to synthesize "tween" frames too. So they can display 120 different images from the 24 film frames. This creates extremely smooth motion. I haven't seen it myself yet and I'm not fully convinced that they can do this without generating too many artifacts. (All the software I've seen that does this generates too many artifacts for my taste. At least with some material.) In any case all those models are hitting the big stores now so we'll be able to see for ourselves soon.