View Full Version : Deinterlace or IVTC? PAL DVD coming from NTSC land
kypec
14th February 2012, 14:42
I never needed anything else besides good deinterlacer for my DVD encoding tasks, well, until now :(
Here you can download my demuxed sample (http://www.mediafire.com/?mgqlgv77bl625xo) together with scripts where you can see what approaches I tried to get decent progressive output.
1. AnimeIVTC = there are still blended frames left behind which I don't like
2. QTGMC followed by selecting every 4 non-blended frames out of group of 5 = result is very good except that offset changes after each major scene change so it's not very usable (there are 10 episodes spread over 3 DVD, ~20 minutes duration each)
3. TFM.TDecimate = basically same results as AnimeIVTC
Can anyone suggest better solution for my footage?
I want to avoid having blended frames like these in the output:
http://i42.tinypic.com/8y6s5y.jpg
http://i44.tinypic.com/2vdh0m1.jpg
Thanks for all your replies!
manono
15th February 2012, 10:30
You should be looking at an unblender. Maybe something like:
QTGMC( Preset="faster" )#or your favorite bobber
SRestore(Frate=25)
But I wouldn't expect perfection from a crap source like that.
kypec
15th February 2012, 16:16
:thanks: manono,
Even with your proposed solution there were still few blends present so I tweaked it a bit further and now it's (almost) perfect! QTGMC(Preset="Fast").SelectEven()
SRestore(omode=2)
Also the speed has improved so what more should I ask for? :D
TheSkiller
15th February 2012, 17:38
With that script you do get rid of most of the blends, yes, but there is one fundamental problem: the frame rate!
The original video was 23.976 fps progressive, the output of that script is 25 fps progressive.
This means there are duplicated and dropped frames in the final output.
I've checked it, the video is very unwatchable like that if you ask me. I'd rather have some blends left over.
You need to remove SelectEven() it's the core problem.
Try this, it gave me good results (and most importantly the frame rate is restored properly).
QTGMC(Preset="Fast")
SRestore(dclip=Bob(), Frate=23.976)
Note that while checking things you should not jump around in the timeline, random access is not supported by SRestore. Always render to a file to check.
manono
15th February 2012, 20:55
With that script you do get rid of most of the blends, yes, but there is one fundamental problem: the frame rate!
The original video was 23.976 fps progressive, the output of that script is 25 fps progressive.
When I suggested my original 25fps script, it was only after carefully checking and deciding that 23.976fps wasn't 100% accurate. For example, the opening live-action guy, at 25fps every frame is unique where there should be a single duplicate every second (or at least a blended frame) were it really supposed to be 23.976fps. The same with the part after that - where the constellation is being drawn in. Every frame is unique at 25fps. As for the main animation, maybe it is supposed to be 23.976fps, but it's drawn, much of it, at 20fps and the jerkiness you see is the same as in any anime, and not because I made it 25fps where it maybe should be 23.976fps. There are (and there are supposed to be) duplicate frames all over the place. However, I agree with you that the moment kpec added SelectEven he ruined any chance of getting it right. I decided this thing is trickier than it appears at first glance and decided it's better to have extra duplicate frames rather than removing unique frames. But it's his video and not mine and he can handle it however he likes. I just threw that script out there to try and set him on the right path - that it needs to be unblended.
TheSkiller
16th February 2012, 00:33
You might be right but the main reason I went for 23.976 was that, to me, it doesn't make sense if the video was originally produced at 25 fps why there are now blends all over place being a PAL DVD. The footage generally looked very much like it has been run through a standard NTSC->PAL fieldblending normconversion box (for example there are those typical Kernel-bob like artefacts as seen in this (http://forum.doom9.org/showthread.php?t=164009) thread, they are however not as pronounced here) which makes me believe it was produced (and maybe edited) in 480i 29.97 via pulldown from what was possibly 23.976 originally.
The footage might be a hybrid, we don't know.
Btw, in my tests I wasn't checking the part at the very beginning with the guy, I was just checking the part where the animated guy walks through the forest with the camera panning with him at whatever full frame rate (every frame has movement).
kypec
16th February 2012, 07:37
Thank you guys for your insightful comments, much appreciated!
@TheSkiller: you might be right about the origins of this footage for I believe too that it's an american (NTSC) title.
Sample clip (same content as my sample :)) can be watched online here at Library Video (http://webcast.mediaondemand.com/library_video/20000901/56_constellation_myths_300.asx) link.
My DVD however comes in PAL flavour for it was produced for local (Eastern Europe) market with dubbed audio. I know I can stretch the audio should I've changed the framerate 25 -> 23.976 but I prefer
to keep PAL framerate if possible.
manono's solution really proved the best once I previewed the clip in realtime playback speed (25 fps).
I was indeed foolish trying to get as few blends as possible because though it looked fine while stepping in AvsPmod the realtime playback exhibited more-than-necessary jerkiness and opening live-action
also felt kinda unnatural to me.
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.