Log in

View Full Version : Almost Every Other Frame Interlace?


Kushami
20th May 2014, 03:15
I'm working with two sets of footage with really bad interlacing problems.

http://twistygadget.com/stuff/lupin/coc_trailer01.png

This footage is 100% supposed to be 24fps, and should be 3:2, however it looks like almost every other frame is interlaced. I stepped through the footage and I can't really seem to make out what's actually going on though. It seems like most of the pieces of the original frames are there, however I'm not sure exactly what the best approach to deinterlacing this would be. I did try some stuff with TFM and didn't have a lot of luck. I unfortunately don't know a ton about the other options, though I imagine something really slow and intensive would be what I want, hahaha.

In the end, I just want to deinterlace it with the least ghosting possible, crop out the black borders, and then scale it down to DVD size.

https://mega.co.nz/#!MZJBTIJZ!FWvDOuz3wo-E8RPTauOUK-c6qxlEAzpeGOSTOBDOERg

Any help appreciated. Thank you very much!

http://twistygadget.com/stuff/lupin/coc_ld01.png

The second set of footage is just as bad if not worse and I can't make out the pattern at all. It's footage I captured from a laserdisc. Looking at it, I did do a separatefields() and stepped through it, and it looks like the footage has ghosting on it no matter what. I have less of an idea where to approach that one than this one. Honestly, I'm not looking for something perfect, just the best case scenario.

https://mega.co.nz/#!lJRTkJLY!ucw5ADzr9Ceu-D-NMuvONh6XgzgC6_UdwwGgUZjFAE4

johnmeyer
20th May 2014, 03:42
I only looked at the first one. It has blended fields. It should be pretty easy to decimate them and then re-time the video to whatever final frame rate you want. Others can suggest what would be the best tool ... SRestore?

Kushami
20th May 2014, 06:49
Thanks for taking a look.

I've been experimenting with Srestore for both and no matter what settings I do I seem to not get any results. I'm not as familiar with it as TFM, however I've tried different omodes (2, the pp ones) and I've not seen any noticeable difference.

turbojet
20th May 2014, 07:56
First sample, this works well:

yadifmod(mode=1,edeint=nnedi3(-2)) #or your favorite bob deinterlacer
srestore(12)

Couldn't download second one.

Kushami
25th May 2014, 08:21
I had to modify the script a little bit, but the script you posted rather well on the first one. Thanks! On the second one, it caused frame studder pretty terribly.

Here's a fixed link on the second one.

https://mega.co.nz/#!lJRTkJLY!ucw5ADzr9Ceu-D-NMuvONh6XgzgC6_UdwwGgUZjFAE4

Looking through the second one, it seems to start as 3:2 but it just gets worse as it goes unfortunately.

Kushami
25th May 2014, 12:49
Not to double post, but I actually got fairly good results here:

AssumeTFF()
Telecide(guide=1,post=2,vthresh=25)
Tdecimate()

There's too much frame blending, I think, to actually try and use something like the above. There's blends 2 frames away in certain points?

turbojet
25th May 2014, 20:39
Second one is 8 fps. Seems srestore doesn't go below 12 fps. This works:

converttoyv12(interlaced=true) # great compatibility or converttoyuy2() for higher quality, slower decode, poor compatibility
yadifmod(mode=1,edeint=nnedi3(-2)) #or your favorite bob deinterlacer
srestore(12)
Tdecimate(cycle=3)

Telecide.Tdecimate caused many mismatches for me last I tested. Telecide causes many problems on near static scenes. Neither of these problems with tfm.tdecimate.

Guest
25th May 2014, 21:12
Telecide.Tdecimate caused many mismatches for me last I tested. Telecide causes many problems on near static scenes. Neither of these problems with tfm.tdecimate. Can you please provide me a stream that demonstrates that? Thank you.

turbojet
25th May 2014, 21:50
Which? Telecide.TDecimate issue happened on most of what I tried. Telecide issue I don't have any specific clips here currently but will look for a scene that will trigger it which might take days or even weeks. I explained a common case at http://forum.doom9.org/showthread.php?p=1658871#post1658871 and tritical replied with how tfm handles such cases correctly.