Restore24 and X-Files R2/PAL

[Morte66]

I've been trying to clean up X-Files Season 1-4 PAL discs, which show pretty horrible blending/ghosting. I had a go at S1E01 with an older version of Restore24 (http://forum.doom9.org/showthread.php?t=75432), which got rid of the nasty edges on motion but sometimes left the video very jerky. I was at a loss to deal with this, and gave up. But reading the Can this horribly field-blended PAL movie be saved? (http://forum.doom9.org/showthread.php?t=105905) thread has inspired me to have another go.

I just bought a used copy of the R1 NTSC set for season 2, and I find when I play it that:
(a) playing the DVDs, Zoom Player keeps jumping between 29.97 and 24fps
(b) in a "straight" backup (Robot4Rip used per the GordianKnot guide, into MeGUI), DGIndex 1.4.5 would keep flipping between NTSC for a few seconds then Film for a few seconds, ending on about 85% FILM overall.
(c) the output of the "straight" encode was clean a lot of the time, with periods of nasty edge combs during motion.

I have no idea whether this is plausible/possible, but the NTSC disc led me to wonder whether the early X-Files might have been shot on a mixture of film and video then edited together. If so, I wonder whether that might have cause the intermittent jerkiness after I used Restore24 on the PAL version.

Do you folks think I might be on the right track?

If so, is there anything I can do about it? Can I get a clean and smooth encode of this material, starting from either PAL or NTSC discs? If so, how would I go about it, and where would I start learning?

[foxyshadis]

You can look at http://www.avisynth.org/VariableFrameRateVideo as well as the readmes included in tritical's TIVTC package to learn a lot about handling of hybrid video. It definitely sounds like that's what it is.

Would it be possible to include a piece or two of the vob that includes a transition to and from video framerates? I'm working on something to do motion compensation for film/video segments, as opposed to the current best cfr solution of blending. Unfortunately I don't have any (or at least know of any) hybrid video in my current dvd collection.

[Morte66]

You can look at http://www.avisynth.org/VariableFrameRateVideo as well as the readmes included in tritical's TIVTC package to learn a lot about handling of hybrid video. It definitely sounds like that's what it is.

Thanks. I'll do that.

Would it be possible to include a piece or two of the vob that includes a transition to and from video framerates? I'm working on something to do motion compensation for film/video segments, as opposed to the current best cfr solution of blending. Unfortunately I don't have any (or at least know of any) hybrid video in my current dvd collection.

Sure. How should I prepare it? Would it be OK to rip and demux the video in DVDDecrypter (straight to .M2V with no DGIndex) then cut with VirtualDubMod, or does it need to be more "raw" than that?

{edit} I could also cut a few minutes out using DVDShrink in Reauthor mode with no compression, to produce a small VOB. Or pretty much anything you like.

[foxyshadis]

Probably a cut with dgindex would be better. (Faster, too.)

[Morte66]

Probably a cut with dgindex would be better. (Faster, too.)

Sample is here (http://www.joelbenford.plus.com/XF201R1NTSC.rar). That's R1/NTSC which showed the rate switching you were interested in.

Now I've seen this in DGIndex, I think some of it has been shot at 29.97 (film or video), and other parts were shot at 24fps on film and had the usual telecine to 29.97, then they were strung together. This would explain why Zoom Player kept switching output rates -- it was playing some footage at 29.97 and doing IVTC to 24fps the rest of the time.

Or, just possibly, the material has been incorrectly flagged as a mix. I remember reading (in software DVD player reviews) that DVD pulldown flags can be very unreliable.

If the R2/PAL discs got frame-blended from that, who knows what happened. Do you want an equivalent sample from the R2/PAL disc?

So, any thoughts on how I might get clean/smooth video out of all this?

[Morte66]

I just used TIVTC on a clip from the NTSC discs to create a timecode file, and muxed it into VFR Matroska. The output is very clean and smooth. I gotta try this with a whole episode, and see if I get any combs.

That worked just great on a whole episode. Smooth motion and no nasty comb effects. The output file was about 10% smaller too (x264 constant quality).

Thank you all for your help, folks.

[scharfis_brain]

you cannot use R24 with the PAL DVD of the X-Files, cause it is
a) sometimes VFR
b) PAL Speedup

R24 will only put out CFR, unless you deactivate the Decimator
the worse thing is the PAL SPeedup.
The X-Files had been converted using a DEFT conversion machine.
THis means most FILM sequences had been IVTCed, so they became progressive. only scenes where the DEFT machine wasn't sure about the underlying content it did fieldblending.
Restore24 isn't able to deal with such nixed contents (25p, 25p blended into 50i, and 31.25p/62.50i blended into 50i)

SO, if you got the NTSC version, do not bother with the PAL one. It is a pain to revert them!

[Morte66]

you cannot use R24 with the PAL DVD of the X-Files
[...] SO, if you got the NTSC version, do not bother with the PAL one. It is a pain to revert them!

OK, looks like it's time to hit Amazon/eBay for used R1 seasons 1/3/4.

[The recent UK R2/PAL discs get better from season 5 when it goes anamorphic -- they look like standard 24->25fps PAL conversions, of about average quality.]

[Mr. Brown]

try this skript from MOmonster

Function unDEFTmatchermod(clip clp)
{
###### PREPARATION ######
global combclip = clp
global matchmask = clp.tfm(mode=0,pp=2,MI=82,clip2=clp.greyscale())

###### VAR.. ######
global match1 = false
global matchc = false
global match0 = false
global matchb = false

###### Conditional Function Chain, evaluated from bottom to top (!) ######
c99=scriptclip(clp, "match1 == true && match0 == true && matchex == true && matchc == matchb && matchc == matchbb ?
\ matchmask : combclip.FieldDeinterlace()")

c5=FrameEvaluate(c99, "global match1 = AverageChromaU(matchmask) != 128.000 ? true : false
global matchc = ChromaUDifference(matchmask,combclip) < 0.001 ? true : false")
c4=FrameEvaluate(c5, "global match0 = match1
global matchb = matchc")
c3=FrameEvaluate(c4, "global matchex = match0
global matchbb = matchb")
return(c3)
}

he posted it in the german x-files thread german x-files thread (http://forum.gleitz.info/showthread.php?t=24901)
95 - 98 % of every episode on the pal DVD's are phase shifted (no problem with tfm)
the rest 5 % (this are only 2 - 3 scenes with a few seconds) are the fu**ing fieldblends (to get this smooth it was processed by Fielddeinterlace())


or 2nd solution
TDeint(order=1,mode=1,full=false,tryweave=true)

excuse me for my bad english