how to deinterlace x-files seasons rc2 ?

[Hobojobo]

I did use the search button and I did read through the guides (Force Film, IVTC, and Deinterlacing- Tutorial from hakko504 etc). Very interesting...
I am not very much into this, I must admit. :(

I want to backup "x-files" Season 2 with Gknot in good quality.

It`s a US-TV production (NTSC), right?
Now it is on DVD as PAL.
Is it than true interlaced, or what?
Can anybody tell, how to deinterlace it ?
Smart or fast or even smart-bob (what?) ?
Perhaps somebody has already done it. Feel free to share your experience.

Thank you.

[Fox Mulder]

What does DVD2AVI tell you? NTSC or PAL? Maybe there is a PAL master despite the fact that is a US tv show, I don't know. You can check that by the resolution; 720x480 NTSC and 720x576 PAL.
Perhaps you have a bad NTSC to PAL trascode, that info is on the forum. If it's really PAL, you don't have to deinterlace, just check Video->Field Operation->Swap Field Order in DVD2AVI.
There is also PAL interlaced, again, search the forum.

[N_F]

Is it a box with 7 (or is it 8?) DVDs? I borrowed this from my brother a few months ago and I've encoded them. This version is PAL. This was a time when I didn't know what fast deinterlace, field deinterace, field shifting and such were, so I just used fast deinterlace. On some scenes you can see some ghosting effects, then again, I'm not sure another deinterlacer would have done a better job, but it's quite likely.

It definitely needs to be modified in some way. How familiar are you with avs-scripts? Try adding decomb's filter Telecide() and look if the interlaced artifacts are gone.

I'm quite certain a deinterlacer needs to be used. In retrospect, I'd probably use Fielddeinterlace (also a part of decomb). But feel free to experiment with others.

[Beave]

This is probably one of the best topics for Fox Mulder!
This whole interlacing stuff is a huge government cover up. They want us to use all sorts of filters...
But I'm no expert here, ask Mulder.

Anyway, I usually use FieldDeinterlace, but which slows down encoding time a lot. I encoded a lot of series and they all were interlaced (Saw those combing effects), even when they were PAL.

If you decide to go for a resolution of 384x288 or below you won't need to deinterlace. I find this resolution good enough. You get more episodes on one CD and it goes a lot faster to encode.

[aramo]

I did the R2 box set a few months back, from my notes I used

Output 512x384 2p125+ mp96@32kHz 16b
Source DVD R2 audio 2ch
Filters: Telecide(guide=2), Neutral BiCubic
Subtitles: Nederlands English Français Greek Italiano
Compiled using P2-300 DivX4.12 GordianKnot0.23b decomb3.91

this gave 2 episodes per CD, telecide comes with the DECOMB addon and is currently at version 4.x

[N_F]

Why Telecide(guide=2)?

Wasn't it a PAL 25 fps version?

[manono]

Hi-

Telecide() without Decimate() (which keeps it at 25fps) can often fix the common "Shifted Fields" problem of PAL DVDs. At first glance they may seem to be truly interlaced, but they are really progressive. From the DecombHelp.html:For simple PAL guidance (tries to maintain lock to the field phase), set guide=2.Fox Mulder's suggestion of using Swap Field Order in DVD2AVI would accomplish pretty much the same thing (and who would know better than he how to work on the X-Files DVDs).

Try adding decomb's filter Telecide()...

And that was from you earlier N_F, so maybe it was just the guide setting you didn't understand.

[aramo]

Beave, I was curious about your comment "If you decide to go for a resolution of 384x288 or below you won't need to deinterlace."

I just tried it with an R2 trailer of 'Withnail & I' which looked to me as if it needed Telecide(guide=2) but I opted for a simple resize to 544x288 [don't see why the width should matter] and the output is combed. I think I follow the logic but does it not require you to keep the matt so that 576>288 loses every other line?

Input = 692x391 [crop before resize]
Output = 544x288
gKnot0.23b divx4.12

as a btw swapping field order did not remove combing, vertical reduce by 2 was no better. It's toss-up between FieldDeinterlace() and Telecide(guide=2) with Telecide having imho the edge. All this is very reminisent of the X-Files S2 down to the sharp-soft alternating frames in some scenes.

[N_F]

No, I'm familiar with the guide setting... I must have been tired when I made that post, I can't make much since of it all either. :)

[Beave]

@aramo

Don't forget to check 4:3 in GKnot. I believe X-Files is not anamorphic. This way you will get 384x288. To get the logic with anamorphed stuff, simply set it to 4:3 resize it down until the height is 288. Just then start cropping and setting the radiobutton back to anamorph. Don't resize after that. The height should be less then 288 then.

I will try the telecide(guide=2) soon as well, since I have lots of TNG DVDs to encode. I hope its faster than Fielddeinterlace. Reduceby2 seems a little too blurry for me.

The X-Files Eps should be quite compresseble, since they are so dark, aren't they?

[aramo]

Beave, you've lost me with the switching in/out of 4:3 and 16:9 - however my previous post refered in part to a movie which had similar combing characteristics to the X-Files. Your right about vertical reduce, I never use it but it might be OK for bonus material.

Hobojobo if you find it helpful here's the AVS I used in conjunction with gknot for the main part of S2E5 aka Duane Barry (Reg2), why not use settings on a short clip which highlights combing and see how it goes. As for time it took 12.5 hours on a P2-300 to get from vob to avi. I recall spending another 30 min or so recompressing the opening X-Files theme sequence [saving about 15MB per episode]

#
# Created with Gordian Knot
#
# http://thewef.nav.to
#
# PLUGINS
# get them from http://users.win.be/dividee
LoadPlugin("C:\PROGRA~1\GORDIA~1\mpeg2dec.dll")
LoadPlugin("C:\PROGRA~1\GORDIA~1\decomb.dll")
#
# SOURCE
mpeg2source("D:\dvd301\xfiles0205\xfiles_0205_43.d2v")
#
# TRIM
trim(0,63925)
#
# IVTC
#InverseTelecine(40,10,15)
# or use
#GreedyHMA(1,0,4,0,0,0,0,0)
Telecide(guide=2)
#
# CROPPING
crop(4,0,700,576)
#
# DEINTERLACING
#SmartDeinterlace(2,15,true,true,true)
# or use
#VerticalReduceBy2
# or maybe
#GreedyHMA(1,0,0,0,0,0,0,0)
#
# SUBTITLES
#VobSub("FileName")
#
# RESIZING
BicubicResize(512,384,0,0.5)
#
# DENOISING: choose one combination (or none)
# 1) little noise (fast)
#TemporalSmoother(2,1)

[N_F]

Originally posted by Beave
The X-Files Eps should be quite compresseble, since they are so dark, aren't they?
Yes, they are very compressable (most of the episodes anyway). I used 3 episodes/CD (that's ~130 minutes fullscreen!)

Originally posted by Beave

I recall spending another 30 min or so recompressing the opening X-Files theme sequence [saving about 15MB per episode]
I wish I'd though of this... Though it must have been very time consuming. Doing something like this for one movie or a few episodes might be ok, but for 100+ episodes (if you do all the seasons that is)?

[aramo]

N_F, regarding time to do of "I recall spending another 30 min or so recompressing the opening X-Files theme sequence [saving about 15MB per episode] "

You could just reuse the same recompressed opening theme clip for all episodes :) noting that "the truth is out there" does vary with some episodes - cliping from beginning of 1st fade-to-white to end of last [or penultimate?] fade-to-white should work well. Fade-to-white is very hard on the divx codec.

Some TV series' are 'opening theme reuseable' friendly - as in it's exactly the same for every episode bar the last few seconds.

[N_F]

I didn't really mean the encoding time (that should be less than 5 minutes on a modern computer), rather the time to cut and paste it manually.

[Hobojobo]

Hi,
thank you @all for your replies.

At the moment I am, sort of, busy.
The resolution 384x288 seems cause the least trouble. :)
The quality is ok for me. I should get 2 episodes on cd.
That sounds reasonable.

Best regards.

[N_F]

You will probably have a hard time filling a CD with 2 episodes if you use that low resolution (the codec will be "maxed-out").

[Beave]

Not if he uses two or more AC3 streams in it ;)

[Slogra]

You can also make an SVCD of each episode. I did it with one episode once and it looks really great. Just keep the episode interlaced at a 480x576 resolution.

[N_F]

Originally posted by Beave
Not if he uses two or more AC3 streams in it ;)
Well, each episode comes with one 192 kbit AC3 stream, so that'd be hard.

[Arkay]

This all depends on the quality you want to achieve. Any of the methods above won't be as smooth as it would look if you played back the original interlaced DVD on an interlaced display as your throwing away some information (2 fields -> 1 frame).

I've done quite alot of interlaced PAL to progressive conversions and the best method I've found is to use smoothdeinterlacer.

This results in a 50fps avi (as each field is converted to a full frame, 2 fields -> 2 frames played double speed), the result of course is babies bum smooth. You're doubling your framerate but it really doesn't impact on avi filesize as the diff between every 2 fields is minimal and the codec is only interested in the differences.

Try this script out (you can get the avs smoothdeinterlacer plugin on the net somewhere just do a search).

LoadPlugin("<path to>\mpeg2dec.dll")
LoadPlugin("<path to>\smoothdeinterlacer.dll")
mpeg2source("<path to>\episode.d2v")
crop(x,y,x,y)
smoothdeinterlace(doublerate=true)
xxxResize(x,y)

I encode all my episodes to 1 700mb CD with ac3 audio (i.e. ~DVD quality) and then burn them 5-6 episodes per DVD-R). This will slow your encoding a hell of a lot too.. I get 10fps on a 2ghz P4. But I figure the end quality is what I want, who cares how long it takes to encode.

Make sure you preview this in vdub before you either manually encode or let gknot encode it. If you get the field order wrong you'll generate a perfectly unwatchable shaky avi (field 2 displayed before field 1). If this happens either modify the field order in your .d2v file or use smoothdeinterlacer(doublerate=true,tff=false) or ttf=true depending on the source.

As for playback my P3 733 has no problems with these files. Anything slower probably won't cope with the framerate.

This only works on truly interlaced PAL material. It'll make Telecined or hybrid material look really bad.

Cheers,
Arkay.

P.S. If you're encoding in Gknot make sure to turn off recalulate bitrate if needed. It'll see double the frames than it expects and halve the bitrate on you which stuffs up your output file size...

[Hobojobo]

Hi, it is me again.
I did some tests...
X-files, pilot
Video: Divx 5.02, two pass, 512x384, Resize Filter: Neutral Bicubic, Telecide (), ca 1000 kbit/s
Audio: Lame 128 kbits abr
size: 324 MB

I tried Telecide() with(out) FieldDeinterlace()
It looks quite promising.

New questions turned up:
1) Is it a good idea to combine these filters, or unnecessary ?
2) What is the difference between Telecide() and ...(guide=2)?

Thanks.

I discovered the preview-function in GK :) and lots more.

[jggimi]

1) Is it a good idea to combine these filters, or unnecessary ?Telecide includes field deinterlacing by default. It is unnecessary. If unneeded, you can turn off the automated field deinterlacing using the (post=false) option.2) What is the difference between Telecide() and ...(guide=2)? It's for field matching pattern guidance. By default, blind field matching is used. Guide=2 is for known PAL content.

You'll find much more on the use of these and other Decomb filters and their options by reviewing the Decomb help document in your Gknot .../docs folder.

(Start...Program...Gordian Knot...Docs)

[N_F]

I may be repeteating myself, I don't remember exactly what I said earlier and don't have the energy to read it through again.

Telecide() should only be used if Telecide(post=false) seems to remove most interlaced artifacts.

If Telecide(post=false) does not remove all (or most) interlaced artifacts then you should use a deinterlacer instead, for instance FieldDeinterlace()


I don't really know why (perhaps it'll screw up some fields or something)... but if you read the decomb document you'll see the author of this program (neuron2) states this point quite strongly.

[hakko504]

Telecide works by matching nearby fields. This may result in doubled or removed fieds. When the original sequence is an interlaced video then there is no superfluous fields to skip, but telecide will detect some as such. This will result in fields being shifted in time as well as perfectly useful information being discarded. Consider this sequence 1t1b 2t2b 3t3b 4t4b 5t5b. Run it through telecide and you might get something like this as output: 1t1b 1t2b 4t3b 4t4b 4t5b. Now if you playback this sequence (on TV) it will look very strange, as certain frames will a) come too early b) repeat them selves c) come too late. Assume the fields in this example contains one letter in the alphabet, in order from A to J, then the new order would be: AB AD GF GH GJ. Not good.

[N_F]

You're a very nice person to have around hakko504 :)

Originally posted by hakko504
Consider this sequence 1t1b 2t2b 3t3b 4t4b 5t5b. Run it through telecide and you might get something like this as output: 1t1b 1t2b 4t3b 4t4b 4t5b.

... but the built in deinterlacer would take care of this (assuming you didn't use telecide(post=false)), though this would result in a deinterlacing of frames that wasn't really necesary, decreasing quality.

Right?

[hakko504]

Right, and you would get a better deinterlacing (more fluent) if you use the correct frames to begin with.

[Beave]

To be on the save side it would probably look the smoothest when we speed it up to 50 fps, right? What would be the fastest way to do it, meaning giving you the most fps in encoding? And ist it necassary to consider the field order?

I read somewhere else, that i.e. the Star Trek TNG DVDs are impossible to deinterlace smoothly. Some frames interlaced tohers not and so on. It was a discussion for SVCD though, where they can't speed up anything, if you want to stay compatable.

[hakko504]

AviSynth internal command: bob()

Avisynth2.0x plugin: Smoothdeinterlace(doublerate=true)
(much better quality, but slower, and not yet compiled for A2.5)

And yes, it is necessary to consider field order, another thing that telecide doesn't, because it wants to reconstruct progressive frames. It purposely destroys field order to create as progressive frames as possible (with the least interlacing).

[thepi314]

Well... I do have some problems with telecide and camera panning ? Looks quite "jumpy" :-(

And I would like to avoid the doubling of the framerate... any tweaks ?

.:Pi314#

[aramo]

Originally posted by thepi314
Well... I do have some problems with telecide and camera panning ? Looks quite "jumpy" :-(
.:Pi314#

I empathise, I recently had this with 'The Wicker Man' R2 and currently I've got it with 'They Shoot Horses, Don't They' again R2. In both cases if I used
SeparateFields()
in the AVS file I could see that some/many fields where made up of 2 or more images, playing thru the separated fields [which doubles number of frames in the output] I could also see a sort of big-step forward little-step backwards progression. Using SwapFields() before SeparateFields() did not improve matters. As the blended fields were not all odd or even, using SelectOdd() or SelectEven() did not help either.

In the case of 'The Wicker Man' it actually looked OK when I played the avi on a standard R2 TV, on a laptop screen is was so-so.

As I said - I empathise :) - but I have no solution.