Questions about progressive/interlaced 480i/576i/1080i MPEG2 encoding

[Mandaman]

I’m currently researching the basics of how progressive and interlaced material is stored in MPEG2 and played back. I already figured out the difference between frame pictures and field pictures, the different flags including the progressive_sequence flag, the progressive_frame flags, etc. I still have some rather technical questions about this, though. I’d really appreciate some answers from an expert! Thanks in advance.

- Is encoding to MPEG frame pictures much more prevalent than encoding to MPEG field pictures? Expressed in percentages?

- When encoding a progressive source (e.g. 25p or 30p) to frame pictures for DVD, the only thing that makes a standard DVD player (with a CRT TV) play this material as interlaced is the progressive_sequence flag that is set to False, right? So theoretically in this case, is the interlacing really done by the player (as there's nothing interlaced in the MPEG-2 encoding)?

- As I understand it, the TFF flags set to True is much more common than bottom field first, right? Are all commercial movie DVDs / HDTV top field first?

- Do flag-reading progressive players/decoders solely rely on the progressive_frame flag in MPEG2 to determine whether they’re dealing with field-based material or frame-based material? So split the frame picture into fields and send these to the player’s deinterlacer (when progressive_frame flag is set to False) or display the full stored picture progressively/weave (when progressive_frame flag is set to True)?

- Will flag-reading decoders in progressive players (e.g. a software player) always decode a DVD/TS-file to 48i/50i prior to weaving when dealing with frame pictures from progressive frames (progressive_frame flag set to True)? We call it a weave, but since the MPEG2 is stored as frame pictures, the decoder would really simply be displaying the full stored picture; there’s really no reinterleaving/weave going on, right? Or is the weaving always done AFTER the decoding part and will a decoder (even in software) always output interlaced sequences as interlaced first?

- Will flag-reading progressive players always treat MPEG2 that consists of field pictures as field-based -> deinterlace?

- Flagged 2-2 pulldown is basically just the progressive_frame flags set to True, right?

- Do progressive sequences contain progressive_frame flags or are they omitted (or maybe required to be 1)?

Thanks.

[mikiem]

"I’d really appreciate some answers from an expert!"

I'm no expert anything, but if any of this helps...

"...encoding to MPEG frame pictures much more prevalent than ... field pictures? "

Actually depends on source. Film is 24 p, scanned at 24 p, so adding pulldown flags more efficient (cheaper) then telecine process as for TV broadcast. Most DV-type footage is already at or near broadcast standards (interlaced etc.), so video normally stays that way.

"When encoding a progressive source (e.g. 25p or 30p) to frame pictures for DVD, the only thing that makes a standard DVD player (with a CRT TV) play this material as interlaced is the progressive_sequence flag that is set to False, right? So theoretically in this case the interlacing is really done by the player?"

A set-top player will output spec (or close to spec) broadcast TV signals -- that's the only thing your TV is pretty much guaranteed to accept. The s-vid output will be interlaced, it's built-into the electronics -- sometimes it helps to remember the set-top player is actually pretty dumb. That said, the player will read the data in the file and follow the directions basically, then (hopefully ) turn that into a spec signal -- it's up to the specific player's innards what it will accept on what kind of disc. THis is a decent source of what's DVD spec: http://www.thedigitalbits.com/officialfaq.html

IN the case of progressive footage, I'd think you'd want to tell any player the truth re: what it's dealing with, for greater compatibility including progressive scan displays. That doesn't say every player will accept it -- a lot of players will not accept everything in the DVD spec.

"AIUI the TFF flags set to True is much more common than bottom field first, right? Are all commercial movie DVDs / HDTV top field first? "

Don't know about HDTV -- not sure sure it's common enough yet to have any general trends... Otherwise no. Set the flag to your source, and really no reason to change it. Mpg2 tends to be tff, DV tends to be the opposite.

"Do flag-reading progressive players/decoders solely rely on the progressive_frame flag in MPEG2 .... split the frame picture into fields and send these to the player’s deinterlacer ... "

From what I've read, & using a bit of logic (trying to anyway ), a progressive scan player outputting progressive, will ivt broadcast spec video as needed, combining fields using some combo of std bob, weave, vector adaptive methods. To me this implies more then pure reliance on flags is at least possible. There is no spec as to how players have to work, there aren't even standards for what they output using component AFAIK, so I'd guess tell the truth unless authoring just for your home TV.

"Will flag-reading decoders in progressive players (e.g. a software player) always decode a DVD/TS-file to 48i/50i prior to weaving ... Or is the weaving always done after the decoding part and will a decoder (even in software) always output interlaced sequences as interlaced?"

For PC display, your software will normally convert any interlaced vid to deinterlaced -- sometimes you can set this behavior. Additionally your graphics card(s) driver might also control display of the video overlay, so it can get complicated -- what do you get if you set the player to interlaced and your ATI drivers to deinterlace? At the same time a lot of video doesn't have flags or any data on interlace or progressive. My guess would be that software playback relies as much or more on content, especially when using adaptive deinterlace.

[manono]

Hi-

I'll take them numerically. There are a few I'll be doing educated guesses on, and may have to be corrected by the likes of edDV or others later on.

1) I've only seen a few encoded as field pictures myself, and only from small DVD production companies using (I would guess) obsolete encoders. I'd say more than 99% are encoded as frame pictures.

2) When you say a CRT TV set, do you mean a standard interlaced CRT TV? Then the DVD is output as fields (480i or 576i). All DVD players originally output 480i or 576i. It's when you have it set for progressive scan and are outputting to a progressive display that an additional step comes into play to make the output progressive again.

You're really asking about 2-2 pulldown, and the majority of DVD players don't handle it well. Most will deinterlace when feeding a progressive display. My experience is with NTSC material mostly. If you're equating the outputting of fields by the player with interlacing, then yes, it's is always done by the player.

3) I don't know what AIUI means. Almost all DVDs are TFF, although there's no rule about it, I don't believe. I see BFF from time to time.

4) In at least one sense everything is field based, as it's all originally decoded into fields. What flag readers rely on most is the presence or absence of the RFF flags to determine if it "IVTCs" or just deinterlaces when outputting to a progressive display. Where flag readers fall down is on the large amount of film based material encoded as already telecined 29.97fps interlaced, either for the whole video, or for parts. They can only deinterlace it, whereas a cadence reader can perform the IVTC to restore the progressive frames.

5) I think you mean 60i/50i. And the answer is yes. The rest of what you said is incorrect, sort of. It doesn't just display the frames as stored on the DVD. It decodes them (if NTSC) at 60i, then IVTCs it (or reweaves it, as you put it) and outputs it in a 3 2 3 2 pattern for 60 HZ NTSC HDTV sets, and as 2 2 2 2 for 50 HZ PAL HDTV sets.

6) Yes, but as mentioned above, there are very few DVDs with a field picture structure. But even frame based DVDs encoding film as 29.97fps interlaced will be deinterlaced by a flag reader.

7) I don't know much about 2-2 pulldown, as I see so little of it. What I have seen is encoded as interlaced, even if the source is progressive. The same is true for PAL DVDs. I haven't seen any where the progressive frame flag is set on a retail DVD (although I set it myself). I believe flag readers will deinterlace it with a slight loss of resolution.

8) They're supposed to have the progressive frame flag set if it's film based. Fairly common is the case where the flag alternates on and off every other frame. You run one of those vobs through DGIndex with the Preview on, and it flips back and forth between progressive and interlaced very quickly. Depending on the player, that can lead to very bad results.

================================================== ================

The main difference between flag reading progressive scan DVD players and cadence reading progressive scan DVD players is that flag readers can only do the equivalent of a Force Film like DGIndex. They can only read the RFF and TFF flags to restore the progressive frames. In their absence they will deinterlace. A cadence reader can read the 3-2 cadence and perform a full IVTC. They are much more robust, and usually more expensive.

I don't know much about transport streams, so I've ignored that. I see that mikiem also answered while I was writing. I hope we didn't contradict each other. I'll read his reply later. I also don't claim to be an expert, but it's a subject I'm much interested in, and if I've made mistakes or said anything stupid, I hope I get corrected. I did the best I could. From the way you were writing, I think you've seen these pages, but for any others also interested in the subject:

http://www.hometheaterhifi.com/volume_7_4/dvd-benchmark-part-5-progressive-10-2000.html
http://www.hometheaterhifi.com/volume_10_1/dvd-benchmark-guide-to-progressive-scan-sho...ut-1-2003.html
http://www.hometheaterhifi.com/cgi-bin/shootout.cgi?function=search&articles=all#Oppo%20D igitalOPDV971H%20(DVI)

Those guys wrote the Bible.

[Mandaman]

Thanks for your input, manono and mikiem. Let's see if some experts can join in to give some definite answers..