Need help ivtcing/deinterlacing a particular video

[aztec321]

So I'm trying to find the best possible way to apply ivtc to be able to watch Freaks and Geeks. I have all the episodes remuxed as mkv and I'm confused as to what exactly I need to do in order to get rid of the interlacing artifacts. They always (I think) occur when there's a change in the camera shot.

Like this one, between shots- http://img856.imageshack.us/img856/74/72435833.png

I've analysed the episodes with dgindex and they're all progressive and >96% film.. So I'm assuming it needs to be ivtc'd. I was previously using my nvidia graphics card to do it all for me, and it did it relatively well, but I want to use madvr instead so I need an alternative. So far I've tried using TIVTC in ffdshow, but playback is choppy. Can someone come up with a better script? Or is there another, more efficient method that I'm missing? Is there a way to do this permanently (as these are the only mpeg-2 files that I'll ever require this for)?

I've upload 3 clips that have a lot of artifacts. They're pretty different from each other, so I hope that helps. If anyone wants, I can upload them to mediafire.

Edit: Fixed links:
http://hotfile.com/dl/111716615/6d52ffa/Twinkies.m2v.html
http://hotfile.com/dl/111716453/b326557/Credits.m2v.html
http://hotfile.com/dl/111716452/848685d/Creepy_sex_conversation.m2v.html

Thanks, all help is appreciated.

[Nightshiver]

You're not going to get much help.

1) Why the hell did you remux to .mkv?
2) To properly fix this, we need source files, and source file's are not .mkv's or .avi's.

Seeing as how this is your first post, I'm going to assume you downloaded these episodes.

[aztec321]

1) Simplicity, I guess. I used makemkv.
2) No, I have the box set. The full DVDs are actually impossible to attain illegally.

So should I split the vobs into clips and upload those?

[Overdrive80]

You should use Dgindex. Open vobs file, select a part & in 'File->Save project and demux video'. It will generate a m2v file, that file must be uploaded by you.

[aztec321]

First post updated with .m2v's.

[manono]

It's fairly standard stuff for TV shows - shot on film and edited on video without concern for the cadence. It's also easy to fix:

TFM(D2V="C:\Path\To\Video.d2v")
TDecimate()
I've analysed the episodes with dgindex and they're all progressive and >96% film.
It's that remaining 4% that's ruining them for you. For example, here's a part of the D2V for your first sample:
900 5 0 0 0 0 0 30 30 d0 f1 f2 e3 f0 f1 e2 f3 f0 e1
900 5 0 503008 0 0 0 72 73 d0 f1 f2 e3 f0 f1 e2 f3 f0 e1
900 5 0 1036675 0 0 0 72 73 d0 f1 f2 e3 f0 f1 e2 f3 f0 e1
900 5 0 1549542 0 0 0 72 73 d0 f1 f2 e3
900 5 0 1799132 0 0 0 30 70 d1 f2 f3 e0 f1 f2 e3 f0 f1 e2
900 5 0 2254524 0 0 0 73 70 d1 f2 f3 e0 f1 f2 e3 f0 f1 e2
900 5 0 2612255 0 0 0 73 70 d1 f2 f3 e0 f1 f2 e3 f0 f1 e2
900 5 0 2956931 0 0 0 73 70 d1 f2 f3 e0 f1 f2 e3 f0 f1 e2
900 5 0 3314727 0 0 0 73 70 d1 f2 f3 e0 f1 f2 e3 f0 f1 e2
900 5 0 3711243 0 0 0 73 70 d1 f2 f3 e0 f1 f2 e3 f0 f1 e2
900 5 0 4096495 0 0 0 73 70 d1 f2 f3 e0 f1 f2 e3 f0 f1 e2
900 5 0 4474771 0 0 0 73 70 d1 f2 f3 e0 f1 f2 e3 f0 f1 e2
900 5 0 4822263 0 0 0 73 70 d1 f2 f3 e0 f1 f2 e3 f0 f1 e2
900 5 0 5164059 0 0 0 73 70 d1 b2 b2 a2 b2 b2 e2 f3 f0 e1
900 5 0 5528824 0 0 0 72 73 d0 f1 f2 e3 f0 f1 e2 f3 f0 e1
900 5 0 5859419 0 0 0 72 73 d0 f1 f2 e3 f0 f1 e2 f3 f0 e1
900 5 0 6168062 0 0 0 72 73 d0 f1 f2 e3 f0 f1 e2 f3 f0 e1
900 5 0 6477409 0 0 0 72 73 d0 f1 f2 e3
900 5 0 6654807 0 0 0 70 71 d3 f0 f1 e2 f3 f0 e1 f2 f3 e0
900 5 0 7073722 0 0 0 71 72 d3 f0 f1 e2 f3 f0 e1 f2 f3 e0
d00 5 0 7503710 0 0 0 f1 f2 d3 f0 f1 e2 f3 f0 e1
900 5 0 7831852 0 0 0 72 73 d1 e3 ff

FINISHED 95.83% FILM

The part I boldfaced with all the '2's in a row, that's one place where they drop to video.

[aztec321]

Sorry, how do I incorporate the .d2v in ffdshow's avisynth settings? I need the script to affect all the episodes. ..I probably should have posted this in the newbie forums.

Thanks for all your help ;x

[manono]

Sorry, how do I incorporate the .d2v in ffdshow's avisynth settings?
I haven't a clue. Perhaps someone else can help. When you said you wanted to do this permanently I thought that meant you wanted to reencode them. It doesn't get more permanent than that.

[aztec321]

Oh, I would, and I tried, but I was unhappy with the results.. Even at like crf=19. Is it even possible to use this with ffdshow in realtime?

[Taurus]

Oh, I would, and I tried, but I was unhappy with the results.. Even at like crf=19. Is it even possible to use this with ffdshow in realtime?
What you mean "unhappy with the results"!?
manono showed you the right way.
If your encodes are not as you expected, try different settings.
x264:
CRF=19 gives me good results with preset=medium ,tuning=grain on your samples.
If you want more flat images try different denoiser/degrainer.
Is it even possible to use this with ffdshow in realtime?
It depends on yor hardware.
Any recent CPU will do this in realtime on SD content. (No netbooks cpus)

[aztec321]

My CPU is adequate, I think. It's a 2.5ghz laptop i5.

I meant I was unhappy with the way it looked overall, not how it was deinterlaced (I deinterlaced it..).

Specifically, I hated how the grain came out.. So I'll definitely try using tuning=grain and perhaps crf=18.5. I assume tuning=grain is a preset- Would I get a better result with specific manual settings? I do not want to apply any denoising or any other filters that mutilate the source. Any other suggestions of specific settings? Encoding time is not an issue, and file size isn't too much of a concern - though I wouldn't want something too large.. Just big enough to appear transparent.

Thanks.

[AzraelNewtype]

Taurus almost certainly means --tune grain unless you're all talking about using a specific GUI instead of CLI x264, but that's the setting you're looking for.

[aztec321]

Okay, so TFM().TDecimate() works great. However, I'm only getting decent (transparent) results with tune grain + crf 16. Any advice?

Thanks.

[manono]

Okay, so TFM().TDecimate() works great.
Yeah, I'm sure it does. But that's not what I suggested. The encoding goes a bit faster if TFM can field match the soft telecined parts by using the information in the D2V file. That's why I suggested:

TFM(D2V="C:\Path\To\Video.d2v")
TDecimate()

Just adjust for your own names and path. It's all explained in the TFM readme.

[aztec321]

You're right, it is slightly faster. I think I'm good now, with crf 16. Thanks for all the help.

[hamsterstyle]

What is the advantage of feeding a D2V file into TFM()

My useage has always been to load the mpeg with teh D2V file, then using TFM with clip2 overrides to fix deinterlacing issues.

Is there a big advantage of using the D2V as part of TFM?

[manono]

Is there a big advantage of using the D2V as part of TFM?
I think so, if the source is a mix of hard and soft telecine. The encoding goes faster and by using the D2V to make the field matches in the soft telecined parts, it keeps TFM from making mistakes and also from deinterlacing frames that might not need it. If the source is pure film/soft telecine you should Force Film it. If it's pure hard telecine there's no real point in using the D2V as it can't help at all. aztec321's sample was perfectly suited for using the D2V.

[mastrboy]

How do one know when it's a mix of hard and soft telecine? (and where can the information about soft telecined flags be found?)

[manono]

How do one know when it's a mix of hard and soft telecine? (and where can the information about soft telecined flags be found?)
You have a look at the D2V (or the film percentage in the Information panel while the D2V is being created, or the film percentage given at the bottom of the D2V). I showed what to look for when I posted the one from aztec321's sample several posts above. If it goes 0 1 2 3 0 1 2 3 (ignore the letters), it's soft telecine (film), and if it goes 2 2 2 2 (or, less commonly, 0 0 0 0), it's hard telecine. Unless it's pure interlace, in which case it might also be 2 2 2 2 or 0 0 0 0. Have a look at the DGIndex FAQ:

http://forum.doom9.org/showthread.php?t=87809

The DGIndex format has changed somewhat since that FAQ was written, but you'll get the idea. Study especially Q. 31.