Log in

View Full Version : Help with Saber Marionette J Again


stpdrgstr
24th April 2009, 19:18
Hi guys.

Some days ago, I bought Saber Marionette J Again DVDs, because I wanted to play with something old. But man, isn't this UGLY...

First, I have been testing deinterlacing methods, and none of them convince me, but of course, I have almost no experience with really interlaced sources.

Also, the image shakes a lot, as if in drugs. I haven't tested any deshaker, but I'd like to know first a way to properly deinterlace this. I hope you can help me clean this :(

Here, I uploaded some demuxes from dgindex. (http://www.mediafire.com/?sharekey=d4f8f27e5ff886be7069484bded33bcdf1fb0b2fc25e394c)

This is what I have until now, someone recommended me using depan, and I added the rest from tcomb on.


LoadPlugin("C:\yatta_6-129\PLUGINS\fft3dfilter.dll")
MPEG2Source("C:\SMJADVD1\FullDisc\SABERJ_AGAIN_D1\project\VTS_06_1.d2v").Trim(0,45279)

temp = last.TemporalSoften(7,255,255,25,2)
Interleave(temp.Repair(last.TemporalSoften(1,255,255,25,2)),last)
DePan(last,data=DePanEstimate(last,trust=0,dxmax=10,dymax=10),offset=-1)
SelectEvery(2,0)
tcomb()
fft3dfilter(sigma=4,sigma2=4,sigma3=13,sigma4=4,bt=4,plane=1)
fft3dfilter(sigma=4,sigma2=4,sigma3=13,sigma4=4,bt=4,plane=2)
dfttest(sigma=0.4)
ediaa()
awarpsharp(10,1,0.5,1).toon(0.2)
spline16resize(640,480,8,2,-8,-2)


ending=MPEG2Source("C:\SMJADVD1\FullDisc\SABERJ_AGAIN_D1\project\VTS_06_1.d2v").Trim(45280,49138).tcomb().fielddeinterlace().fft3dfilter(sigma=4,sigma2=4,sigma3=13,sigma4=4,bt=4,plane=1).fft3dfilter(sigma=4,sigma2=4,sigma3=13,sigma4=4,bt=4,plane=2).dfttest(sigma=0.4).ediaa.awarpsharp(10,1,0.5,1).toon(0.2).spline16resize(640,480,8,2,-12,-2)


final=last+ending

gradfun2dbmod(final,thr=1.4,thrc=1,str=0.8)


I'm doing sepparated deinterlacing for the ending because the background shows combing, but the credits don't.

Also, DGIndex says the frames of the episode are encoded as Interlaced, but it seems they're rater progressive... but when I check the created lossless from this script with DGAVCIndex, it shows problems to decode it and says it's interlaced. Don't know what to do :(...

Any suggestions, corrections?

*.mp4 guy
24th April 2009, 20:59
Your source apears to be 100% progressive, except for the credit sequence, which is hybrid progressive 29.97P, and telecined progressive 23.976, I would recomend just leaving everything alone, and putting the credit sequence through mcbob and calling it a day. Your source also suffers from dotcrawl, but has already had it mostly removed by a fairly good and sharp 3D comb filter, its your call wheather or not to risk using a comb filter on it, but I would definately recomend that over demolishing your source with fft3d and dfttest and AA.

stpdrgstr
24th April 2009, 23:14
Well, I'll pass mcbob on the credits then and see how it looks, thanks :)

What comb filter do you recommend for this? I'm already using tcomb, if that's what you mean. And I don't know if I should remove AA, this suffers from a lot of aliasing :(

:thanks:

TheRyuu
25th April 2009, 00:16
Not sure what was in the minds of the people who made this, but it would've been a lot better to just make the background animation on the credits 30fps just like the rest of it. :confused:

Anyway, you have 2 choices with the credits... you can either just leave it alone and run it through some type of deinterlacer/bobber (try nnedi and tempgaussmc). Or run tfm().tdecimate() on it and then run nnedi over it afterwards to clean up the scrolling credits. Now one will probably produce a jerky background, the other will produce jerky credits.
(there's actually a third option of using some mvtools and other magic but I think your background is too high motion for that to work correctly here)

It's up to you which one you want to do.

Furthermore, you're probably overfiltering your source with calling fft3d and dfttest and all the other stuff as well so just watch out.

stpdrgstr
25th April 2009, 02:47
Well, fft3d is only removing raimbowing, as far as I can see. That's one of the methods I've found to be really successful at that.