View Full Version : Can't figure out deinterlacing solution
manono
20th February 2010, 20:47
It's only the first GOP (half a second long), either some black frames or the first part of that Board Of Film Censors form. There's no need to split it.
Guest
20th February 2010, 20:50
It's only the first GOP (half a second long), either some black frames or the first part of that Board Of Film Censors form. There's no need to split it. Right. In DGIndex, after loading your file(s) just move in by a GOP by hitting > and then set the project to start there by hitting [.
EuropeanMan
20th February 2010, 20:56
^ thanks guys - then I won't have to worry. The current lag should be fine if the ONLY thing I have to worry about is the CENSOR CERTIFICATE or some other opening credit BS.
I'll assume correctly then that the ASSUMETFF won't affect the rest?
manono
20th February 2010, 21:04
I'll assume correctly then that the ASSUMETFF won't affect the rest?
It had better affect the rest. That's why you put it in there, so Bob/SRestore will work correctly. Didn't you check all this before starting a (what?) 12 hour encode? Don't worry, it'll turn out fine. It's that first GOP that would get messed up if it wasn't just a few static frames. It would have the dropped frames and the dupe frames.
EuropeanMan
20th February 2010, 21:57
^ close to 24 hour encode LOL
Usually don't have to check all of this with most sources I have...this was one of those few exceptions...which was WHY I came here with this thread in the first place.
EuropeanMan
1st March 2010, 22:43
Okay guys - different source.
http://www.sendspace.com/file/vota6d
All indications tell me this is hard-telecined.
interp = nnedi(field=1)
deint = tdeint(mode=0,order=1,field=1,edeint=interp,slow=2,emask=TMM(mode=0,order=1,field=1))
tfm(mode=3,order=1,clip2=deint,slow=2)
tdecimate()
crop( 12, 80, -6, -94).spline36resize(720,304)
I've tried this...which looked better than tfm(order=1).decimate()
But I still get artifacts and blends/ghosting (?) such as this:
http://i45.tinypic.com/208d1k8.jpg
This frame is 173 after using code above. Also used cpu=6.
Any suggestions please? Thanks
PS - I've tried using srestore to 23.976, and it didn't work as I thought it would. 25fps didn't make sence here either.
Didée
1st March 2010, 23:09
Too lazy to count closely, but it looks pretty much like a sort of 24p --frameblending--> 25p --pulldown-->29.97i.
That, or something similar. IVTC to 25 fps is all you can do. I see no way to fix this. It's broken.
EuropeanMan
1st March 2010, 23:15
^ When you say, broken...what does THAT mean?
(now to research a way to IVTC to 25fps) <--thanks, will try that
I'm thinking of WHY this source is sort of double-field blended?
EDIT: is there way to reverse somehow what you've analysed?
29.97i ...> 25p ...> UNframeblend ...? 24p?
Didée
1st March 2010, 23:29
On a 2nd thought, perhaps it was filmed not at film rate but at video rate before blending, that's also possible. Whatever the exact process has been, the blendings are not correlated to the neighbors. Sequences like A B CD E F. You can't unblend CD because the neighbors are not related, the blend is a unique state. Also there are sequences like A BC DE F. That's (impossible)² to restore.
On the top, it's just telecine. Under the telecine, you get what you get, there is nothing that could be done. Unless you come up with a method that reckognizes not-blended frames, motion-compensates those onto the blends, then uses that for unblending. (That's a joke. It is thinkable, but not doable.)
edit - after a fag - yep, probably it's 50i ->blended-> 24p -> telecine. Dead road.
EuropeanMan
1st March 2010, 23:43
LOL - thanks...
but you said "broken" earlier...is this what you meant by that word?
Didée
1st March 2010, 23:44
Stephen, deleting posts is not a nice habit. Especially when done regularly. A moment ago, there was interesting information. The next moment, it is gone.
Happens a lot latey. Not nice.
but you said "broken" earlier...is this what you meant by that word?
Hmh, using the good word is ... not good?
How about "FUBAR". Easier to understand? :D
EuropeanMan
1st March 2010, 23:48
This movie was a FESTIVAL film...so it makes sense that you thought it could have been at VIDEO rate...which is WHAT exactly? Filmed with a handy-cam in some sections, and regular movie camera in other scenes? I ask these questions to LEARN...thanks again.
manono
2nd March 2010, 09:57
Stephen, deleting posts is not a nice habit. Especially when done regularly. A moment ago, there was interesting information.
I agree. He sometimes makes very helpful posts and then removes them. I tried to PM him a couple of days ago to ask for the reasons for this peculiar, paranoid even, behavior, but he doesn't accept PMs. As a mod, I can read them after they've been deleted and, on the face of it, there's no apparent excuse that would justify removing them. Do you have anything to say for yourself, Stephen?
manono
2nd March 2010, 10:57
About all you can do is bring it back to 24.975fps:
TFM()
TDecimate(Mode=0,Cycle=6,CycleR=1)
After that it's hopeless.
EuropeanMan
2nd March 2010, 22:31
^ would using CDeblend() help at all along with the above?
manono
3rd March 2010, 02:35
No, it was deinterlaced and is progressive after the IVTC. Both fields of each frame are the same. A deblender, no matter what it is, depends on having clean fields, not frames where both fields are identical and already blended.
vBulletin® v3.8.11, Copyright ©2000-2025, vBulletin Solutions Inc.