NTSC Encode, Combing Even After IVTC

[Master Yoda]

I have made a backup of The Temple Of Doom R1 version, DGindex said it was 99%film but instead of using force film i used Telecide(guide=1),Decimate(cycle=5), the reason for this was because i read that unless dgindex says it's 100% using force film you may come across the odd combing. But for some reason the new encode has a couple of combing lines 17 frames to be exact but its not at the part where dgindex switched from film to 99%film it has happened much earlier where it's progressive film. Here are two screen shoots, the to m2v's have been run through dgindex first with force film on to create the samples.

After Encode (http://i120.photobucket.com/albums/o176/jedi_master_yoda/Untitled-2a.jpg)
Original (http://i120.photobucket.com/albums/o176/jedi_master_yoda/Untitled-2b.jpg)

Any ideas what went wrong ?? dgpulldown1011 was used to add the pulldown to the new encode

[Guest]

Field matchers are not perfect. The postprocessing should clean up the stray combs. You can use manual overrides too if you want.

But anyway, you should use Force Film for 99% film. Then add FieldDeinterlace(full=false) if there are any stray combs.

[Master Yoda]

I see, i assumed that force film and ivtc done the same thing. Is it safer to just add FieldDeinterlace(full=false) when using force film if dgindex doesn't say it's 100% film ? just in case there is the odd combing.

[Guest]

Force film is a form of IVTC. It works by ignoring the RFF flags. Doing external IVTC with Telecide/Decimate works differently. The RFF flags are honored and then lost and then Telecide/Decimate has to try to figure out which fields are duplicates. It cannot do it perfectly.

My suggestion is that when film % is greater than 95%, do Force Film. If stray combs are seen, then add FieldDeinterlace(full=false) with an appropriate threshold (or use the equivalent tritical tool).

Often the credits are in video and the film itself is all 3:2 pulldown, and this results in a film percentage less than 100%. You need to analyze what you actually have rather than try to apply a fixed process.

I have seen broadcast transcoders taking hard pulled-down film as input work by adaptively detecting video versus film sections. For video, they encode interlaced; for film, they encode progressive with RFF pulldown. Even for a pure film, the adaptive process sees some sections as video and the result is a stream with some hard telecined parts (detected as video by the transcoder) and some soft telecined parts (detected as film). You have to look for this using Parse D2V to see breaks in the 3:2 RFF pattern.

[Master Yoda]

Force film is a form of IVTC. It works by ignoring the RFF flags. Doing external IVTC with Telecide/Decimate works differently. The RFF flags are honored and then lost and then Telecide/Decimate has to try to figure out which fields are duplicates. It cannot do it perfectly.

My suggestion is that when film % is greater than 95%, do Force Film. If stray combs are seen, then add FieldDeinterlace(full=false) with an appropriate threshold (or use the equivalent tritical tool).


I see, problem is if the film is 95% or higher i can't site there and watch all the way through to see if there's any stray combs and i don't really want to have to do a new encode if when i watch the movie afterwards i see a stray comb that's why i ask if it's save just to add FieldDeinterlace(full=false) to be on the safe side. Reading the decomb manual it seems to me at least that adding that line wouldn't do anything to the video unless it detects interlacing.

[Guest]

There's no universal prescription. The results will be proportional to the time you're willing to invest. That is all I have to say.

[Master Yoda]

OK Cheers.