I am trying to encode a dvd that I purchased lately. But there is weird artifacts (or whatever that is) on pattern clothings (the teacher's, see the sample). Plus, some parts of the movie get shaky.
So I wonder how I should approch this movie. It's seems to be telecine probably mix with something else. But I am really a noob when it comes to analyse a video or their problems.
If anybody could give me some guidance on this I would really appreciate.
So far I use this script with avisynth to deal with the telecine material
Gives me good results with the clothings parts (some are left, I might have to deal with them manualy?). But I don't know how to deal with the shakiness, which is present on the dvd itself. Any other filter that I have tried like yadif, qtgmc, or other TFM setting makes the clothing part get crazy...Mpeg2Source("the source")
I have attached a sample...
Thank you for your help
+ Reply to Thread
Results 1 to 5 of 5
TFM's post processor (which looks for residual combing after field matching, and when found, deinterlaces it) is sometimes confused by alternating thin horizontal lines like that. You can disable the post processor by adding the pp=0 parameter to TFM(). But that may let some real combed frames through (usually an orphaned field at a scene change; I didn't notice any in this clip). Or you can try changing the cthresh parameter to make it less sensitive. Changing it to 40 kept it from kicking in. But when set that high it may also let real combed frames through. One other way of working around the problem is to provide a separately deinterlaced clip of your own to use when residual combing is detected. But other deinterlacers will have problems with these shots too so that probably won't work.
Thanks for your answer Jagabo. I will try to tweak TFM's setting a little bit to see what kind of result I get.
But when it comes to shakiness, like when you look at the teacher's desk at some point or the girl's book on the bottom. Is there something to do with this ? I tried Stab() but I don't see much difference...Any filter to suggest ?
Unfortunately, those are rolling shutter distortions (the film was moving a little as it was scanned) so some parts of the frame move up, some parts down. That's why stab() doesn't help much. QTGMC(InputType=2) helps a little but it messes up the teacher's jacket in those shots where TFM screwed it up.
All right... Thanks, at least I will not spend more time trying to solve this problem.
I really appreciate your time and help