"Colour delay" on IVTC'ed hard telecine

[Skauneboy]

First poster here. First of all I'd just like to say what an amazing forum this is. I've spent the last couple of weeks learning the basics of Avisynth and h264 encoding to backup some of my DVD's and in this process Doom9 has been a vital source of information. So far I have been able to handle the different film/video types I've run into while archiving my DVD's and luckily it's mostly been interlaced PAL but also some soft and hard telecine. Of course eventually you will run out of luck... and this time I haven't found the solution even here on the forum.

The troublesome source is hard telecined and shows symptoms of something I can only describe as "colour delay" (chroma shift?). After being IVTC'ed, every fourth frame has red/yellow and blue/green fields from the previous frame. Pay attention to the guy's face and cuffs:

Source



TFM(d2v="C:\whatever.d2v",slow=2,pp=0).TDecimate()



SeparateFields



Thankful for any suggestions on how to deal with this. Let me know if you want me to upload a source clip.

[Guest]

Where does the source come from?

[Skauneboy]

It's a Korean NTSC R0 from 2001. Any hope for it?

[Guest]

If the separated *source* fields look like you posted, I fear not. But maybe someone has a more optimistic view.

[Skauneboy]

Yeah, I suspected as much. Thanks anyway. So purely out of interest, what would cause the fields to have that discoloration? Only the top fields are affected by the way.

While I have your attention, in a telecined source what is the third field in a frame group with three fields? I mean like this (for TFF): TFBF, TFBF,TFBFxx. I'm hoping it's a duplicate of the top field.

[Guest]

Originally Posted by Skauneboy

Yeah, I suspected as much. Thanks anyway. So purely out of interest, what would cause the fields to have that discoloration? Only the top fields are affected by the way.

I don't know. It could have been created from an old VHS tape source, where such problems are common. Frankly, I deal only in very clean sources (for shows, demos, etc.), so perhaps someone else knows more about the pedigrees of these pathologic sources.

While I have your attention, in a telecined source what is the third field in the third group of frames? I mean like this (for TFF): TFBF, TFBF,TFBFxx. I'm hoping it's a duplicate of the top field.

TFF 3:2 looks like this:

A A B C D # top fields
A B C C D # bottom fields

The two top A fields are duplicates and the two bottom C fields are duplicates.

[Skauneboy]

Excellent, then I should be able to get rid of one bad field in each frame group if I'm understanding this correctly. Time to bring out the TFM documentation.

[Skauneboy]

I have no idea how or why, but a simple nnedi2() removed 99.9% of the colour ghosting.

nnedi2()
TFM(slow=2,pp=0)
TDecimate()

[Skauneboy]

Thanks Stephen, that explains it. I ended up using an override file so that only the affected frames are replaced with the nnedi'd ones.

cleaned=last.nnedi3(nsize=1)
TFM(slow=2,pp=2,ovr="override.txt",ovrDefault=1,clip2=cleaned)
TDecimate()

override.txt

field = top

0,7323 ----+
7414,55695 +----
55785,74347 ----+
74348,108943 +----
109013,112190 +----
112191,127123 ---+-
127214,151216 +----
151326,190168 --+--
190169,198208 ---+-