How to fix color separation after IVTC?

[Zap250]

I have vob files that were telecined to 29.97 fps, so two out of every five frames were artificially interlaced. Entering my IVTC function -- Telecide();Decimate(cycle=5) -- solved the problem but resulted in a hideous blue ghosting effect where the subsequent frame would have been. It's noticeable throughout the video in one out of every four frames after IVTC, particularly when there's movement. Any advice on how I might correct this? Following is a frame capture so you can see what I'm describing:
http://img177.imageshack.us/img177/1811/afterivtctc4.png

And here is the frame sequence before IVTC, in case it helps:
http://img404.imageshack.us/img404/7588/4119cm3.th.png (http://img404.imageshack.us/my.php?image=4119cm3.png)http://img145.imageshack.us/img145/6471/4120pk0.th.png (http://img145.imageshack.us/my.php?image=4120pk0.png)http://img294.imageshack.us/img294/543/4121re8.th.png (http://img294.imageshack.us/my.php?image=4121re8.png)http://img404.imageshack.us/img404/5389/4122pd9.th.png (http://img404.imageshack.us/my.php?image=4122pd9.png)

[cestfait]

chromashift plugin, maybe?

[thetoof]

That'd be my first guess... but it's kinda weird to have that blue ghosting appearing out of nowhere. Could you post a 30-40 frames .m2v (load your vob in dgindex, select a small part and demux video)

Maybe telecide is creating this... dunno how though. You could try using tfm(d2v="path&filename.d2v", slow=2) + tdecimate() (from the TIVTC package)

[nonsens112]

Zap250

did you try setting "forced film" in DGIndex?
Video -> Field operation -> ...

[Zap250]

@thetoof: Tried that tfm function just now and instead of the solid blue ghosting we have interlaced blue ghosting, on 1 of every 4 frames as before. Not sure if that's any better. Here's a frame grab:
http://img90.imageshack.us/img90/8077/clipboard1tt0.png
And here's that m2v sample you requested: http://www.mediafire.com/?yetymy021il

I'm looking at the untouched frames very closely and I don't think the IVTC is not at fault. If you look closely the separation is visible in the interlaced frames but not the progressive frames. Apparently this is a result of sloppy telecining. If I try to fix it with chromashift, of course I would only want to apply it to the frames that have the problem. Problem is, I can't just select every 4th frame since the sequence misses a frame every few seconds.

@nonsens112: Thanks for the suggestion, it still leaves the interlaced frames in place. I need to IVTC it, but without the video getting messed up.

[nonsens112]

just tried
DGDecode_mpeg2source("E:\New Folder\VTS_01_1.demuxed.d2v",info=3)
ColorMatrix(hints=true,interlaced=true)
edeintted = nnedi(field=-2)
TDeint(edeint=edeintted, mode=2, mthreshL=2, mthreshC=2)
TDecimate()
on your sample, seems to be OK.

[nonsens112]

or:
DGDecode_mpeg2source("E:\New Folder\VTS_01_1.demuxed.d2v",info=3)
ColorMatrix(hints=true,interlaced=true)
edeintted = nnedi(field=-2)
deint=TDeint(edeint=edeintted, mode=2, mthreshL=2, mthreshC=2)
tfm(d2v="E:\New Folder\VTS_01_1.demuxed.d2v",clip2=deint,order=-1,Mode=5, pp=6, slow=2)
TDecimate()

[Zap250]

@nonsens112: Thanks for the code - it works great!

[nonsens112]

thanks to tritical :rolleyes:
TDeint, TFM and TDecimate are very usable things, just read the manual and adjust parameters.

NNEDI is another nice thing, but it is quite slow, so you can try
edeintted = SeparateFields().EEDI2(field=-2)
instead of
edeintted = nnedi(field=-2)

[Blue_MiSfit]

Yeah, learning how to _really_ use TFM / TDecimate / TDeint correctly is quite a process for most. I'm still just barely capable of doing IVTC on my own (when it's not standard 3:2)

~MiSfit