Mixed Progressive and Interlaced Encoding?

[jdobbs]

I've searched through the threads and I know this has been discussed, but I haven't found a good answer. Specifically I'm working with the ST-TNG disc set. The episodes are encoded partially progressive and partially interlaced. Not the "constant shifting" type, but portions lasting quite some time in either form. I encode with CCE and use SmartRipper, DVD2AVI, AVISYNTH.

1. Feeding from DVD2AVI with FORCE FILM causes the interlaced sections to be jumpy.

2. Feeding from DVD2AVI without FORCE FILM and using DECOMB with the following parameters:

telecide()
decimate(cycle=5)

-- Same problem, the interlaced sections are jumpy

3. Fed from DVD2AVI without FORCE FILM and no filtering:

The interlaced sections look fine, but the progressive sections get jumpy

4. Fed from DVD2AVI without FORCE FILM and:

telecide()
decimate(mode=1, threshold=50)

Again the interlaced sections look fine, but the progressive sections get jumpy.

I've also tried GreedyHMA in multiple configurations with similar results.

Has anyone come up with a solution for this?

Thanks,
jdobbs

[Eyes`Only]

For videos like that, i use the decomb.dll plugin for avisynth. It supposedly will find the interlaced parts and treat them correctly, as well as ignore the progressive parts, since it's a rather intelligent plugin and knows not to mess with those.

Weird that you're having that problem. Maybe the decomb telecide settings you are using are incorrect? I know theres many more actual parameters that decomb can use to correct certain types of video. Doom's decomb guide shows this recommendation for hybrid video:

LoadPlugin("C:\PROGRA~1\GORDIA~1\mpeg2dec.dll")
LoadPlugin("d:\decomb.dll")
mpeg2source("D:\lain.d2v")
Telecide()
Decimate(mode=1,threshold=50)

[jdobbs]

Yeah, and of the many settings I tried I would say that one looked the best -- but it still had annoying jumpiness... maybe I'm too picky. My only other option would be to only put two episodes on each disc rather than the original four. I hate to do that though -- I'd much rather meet the challange and find a way to beat this. My guess is that we are going to see a lot more of it (the manufacturers have to know this makes copying a pain)... I certainly am pleased with the quality and think the set is worth the $89 I paid for it at BJ's (7 DVDs) -- but I want to make a backup.

[smoof]

I'm assuming its NTSC, so did you try:

Telecide(guide=1) -- (this is not the default)

Alternatively, try replacing the Telecide/Decimate combo with FieldDeinterlace()

Could also be a field order problem that you could switch using pulldown.

My approach would be

1. DVD2AVI force film off (using a small clip that has both prog and interlaced)

2. avisynth
Telecide(guide=1)
Decimate(cycle=5)

3. CCE - settings
progressive frames
top field unchecked (see other posts about this CCE bug)
zig zag
nonlinear
dvd compliant

4. pulldown src dst -prog_frames p

The prog_frames p can make a difference and I always use it for progressive source material.

5. If video is still jumpy, run pulldown again and switch the field order. Be sure and run it on the destination mpv from step 4.

[jdobbs]

Thanks smoof -- I'll try it tonight!

[jdobbs]

Cool! I did a test section and it looks like FieldDeinterlace() alone gives good results. I'm doing a complete episode to see how it looks. Thanks again, smoof... I'll let you know how it comes out.

[TRILIGHT]

I've always used Telecide(guide=1) and Decimate(cycle=5) with good results as well. I can't say I've ever seen anything so seriously split between progressive and interlaced like you have though, Jdobbs. I'd be really interested in hearing what your final results are!

[jdobbs]

Well, I'm still seeing jumpiness, although not as pronounced, on the final encode for those sequences that were originally progressive. I'm going to burn to DVD-R and see how noticable it is on a standalone.

What makes this unique from the typical ones where Telecide(guide=1) and decimate(cycle=5) works is the length of the interlaced portions. Often you won't see the affects of removing the frames because the sequences are so short (they pop into interlaced, then immediately pop out). In this series, you have complete scenes that remain interlaced, and you have the affects of trying to inverse telecine a pure 29.97fps interlaced source.

You would think that feeding and encoding at 29.97 would fix the problem, you'd lose the efficiency and quality of a progressive stream because you encode the result of the pulldown. But it just seems that the inserted frames (from the pulldown) seem to be really pronounced -- especially in high motion or panning scenes.

I wish here were a good way to reencode in the exact same pattern as was originally encoded on the source. I can do that with ReMPEG2 -- but ughhh, what a quality drop.

Well, maybe it's just the way it looks on my PC -- I've seen that before. The disc is burning while I type, hopefully it will look better on my standalone. I'm not feeling optimistic, though.

BTW. Just for info, DVD2AVI sees the source as being about 59% film and 41% NTSC on the episode I just ran it against.

jdobbs

[mikeathome]

Hi,

complete different approach:
- open the *.lst file (List of Vobs) in VStrip
- demux audio
- demux video WITH split by Cell-ID (split at the chapters)
- it's pretty usual that the Field Settings change at the chapter (Cell ID) borders

I had something similar once with a PAL title, where the field order changed from time to time with the chapters (Cell-IDs). To re-encode at correct field order you have only ONE chance in this case: Handle (encode and patch) the different field order parts separately.
Joining the parts is not a big deal, just put them on the timeline of Maestro in the appropriate sequence and underlay the separately grabbed audio, works great...

mike

[jdobbs]

Just an update. It didn't look any better on my standalone. I've since tried and succeeded in recreating the DVD with ReMPEG2. It works completely -- but the quality drop was more than I could stomach.

I did a quick look at the possiblity of the changes happening on cell boundaries a couple of days ago (just scanning with VSTRIP). It didn't look like it fell there -- but I think I'll try anyway.

What we need is a high quality MPEG2 to MPEG2 transcoder! Something that works like ReMPEG2, only that feeds another encoder (maybe the CCE Premier Plugin or TMPGEnc)...

Oh well. At least I got to watch some STNG episodes this weekend while my computer was rolling along doing useless encodes...

[Trahald]

yeah.. i high quality rmpeg2 would be great.. but its such an undertaking (would have to be multipass capable and with rempeg2s current code would be reeeeeallyy sloowww).. slower that is.. hehe

anyways.. u should be able to just keep it interlaced (29.97fps).. its makes things alot harder for cce (and double wammy for film source means your encoding extra frames) but at least playback smoothness should not be affected..

[dvdRENEGADE]

jdobbs,
I left a reply for you in the other thread.

http://forum.doom9.org/showthread.php?s=&threadid=29561