I have a video source that I believe is the result of poor deinterlacing followed by a telecine 3:2 filter. Running an invert telecine through AviSynth using TIVTC is not a major problem using this code:
Now however I am left with a video that shows combing due to the poor deinterlacing... Is there any thing that can be done? I've tried two things:Code:TFM(mode=5, pp=0, slow=2) TDecimate(cycle=5)
1. Handbrake Decomb filter, which slightly reduces the combing but also results in a loss of quality.
2. QTGMC with InputType=1 for progressive sources. Better quality but combing is either gone or enhanced depending on where you look.
+ Reply to Thread
Results 1 to 9 of 9
Post a sample of your source.
I considered doing that but backdashed out of it because it's NSFW, and I don't know what the policy is for this. I can find a non-explicit part but... you know
It's hard to say from just that image. There's no combing, just aliasing. Try an antialiaser like Santiag() or AAA(), followed with a very mild aWarpSharp() to resharpen. Or maybe QTGMC(InputType=2) or 3.
[Attachment 49070 - Click to enlarge]
If you want to make a clip available privately upload it somewhere like mega.nz and PM me a link.
You're right it's more aliasing than combing, my bad. I wonder how it was introduced in the video.
I've tried running Santiag, which is giving fairly good results but aWarpSharp seems more destructive than it is worth. Did you mean running QTGMC *after* Santiag or as a replacement?
No, I meant to try using QTGMC(InputType=2) instead of Santiag(). That setting is meant to clean up bad deinterlacing. But I don't think that's really the problem. It looks to me like the video may have been downscaled with a nearest neighbor filter.
I see, thank you. One final question, is there no way to setup QTGMC for invert telecine? It should only be a matter of deinterlacing certain frames, but I don't see any option for that. Not that TIVTC is doing a bad job, but I'd rather rely on QTGMC.
You can use QTGMC to inverse telecine but it's not as good at it as TIVTC. And it's much slower.