View Full Version : What's the best way to de-interlace this film and Convert To PAL?
Kyaneos
26th March 2006, 16:08
Hi!
I have an anime movie NTSC and cannot de-interlace and convert to PAL (Progressive, 25 FPS and 720x570). I have tried Telecide, TomsMoComp, TDeint, Decimate... and I cannot de-interlace well and to eliminate the effect ghost.
Sample of the original Vob:
http://www.rapidsharing.com/707a977f665411324fbe64e2fd8e12b9
This is the info of the DGIndex about the sample:
http://img20.imageshack.us/img20/6830/temp51il.jpg
This is the info of the DGIndex about the Full Movie:
http://img129.imageshack.us/img129/4786/temp8pf.jpg
Bye!
MOmonster
26th March 2006, 16:23
:readrule: There is no best way.
I canīt have a closer look on your sample here, but there are some thinks you can do. You can use repal or restore24 or Cdeblend + decimator.
Mug Funky
26th March 2006, 16:44
DGindex says there's at least some film in there, but not much. that means if you treat it as film (simple 3:2 pulldown), you'll end up with jerkyness.
i find with this stuff, to get it to PAL with a minimum of fuss is to:
leakkernelbob(order=1,threshold=2)
lanczosresize(720,576)
blendfps(50)
assumetff()
separatefields().selectevery(4,0,3).weave()
to make this work you'll have to go find Clouded (mg262)'s "motion" package, and leak's optimised kerneldeint package. otherwise you can replace blendfps with "converttoyuy2().convertfps(50)" and get the same visual result with some lost precision and speed.
this isn't _the best_ way to go NTSC->PAL anime, but it is _the most robust_ way :)
you'll have to encode as interlaced, top-field first. i also assume your anime is top-field to begin with (99.99999999% of it is, but often extras can be in bottom-field due to stuff being shot on DV). if you get stuttery motion and weird leftover combing (different to the combing introduced by the conversion that is, and far uglier), then you should change "order" to 0 in leakkernelbob.
just FYI, the above method is what i use for all my PAL conversions except for talking head stuff where it's safe to use the newer motion-compensated approaches. also note that the resizing to 576 isn't completely accurate - it should be 480*625/525, padded with black to meet 576 lines. but let's not split hairs... the lanczosresize string for doing that in 1 move is quite long.
scharfis_brain
26th March 2006, 16:45
upload the sample somewhere else. I cannot DL it.
Kyaneos
26th March 2006, 18:24
upload the sample somewhere else. I cannot DL it.
Sample Mirror:
http://s1.ultrashare.net/hosting/fs/903a2795286d4a78/
scharfis_brain
26th March 2006, 18:44
I'd prefer my manual deblending approach on this sample:
i=mpeg2source("C:\Downloads\forum\kyanos\Temp.d2v")#,cpu=6)
i
tdeint(mode=1,full=false,type=3)
changefps(25*10)
selectevery(10, x)
removedirt()
bicubicresize(width,576)
the only thing you need to alter is the x in the selectevery line.
x ranges between 0 ... 9 and should be altered until the motion is smooth and all blends have gone.
that MPEG video has been encoded progressive, but its content is interlaced, so you will notice a chroma lag (or jerkyness), which is not possible to restore.
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.