Hard time to deinterlace old anime DVD

[Shinobu]

Hello.

I'm used to avisynth for all my DVD, bluray filtering.
I use lot of functions of this great forum as i'm not able to do them myself.

Yesterday i found one of my old DVD and watch it again, the quality is realy bad, specialy the interlacing :mad:.

I tried to deinterlace it myself with animeivtc, tdeint, it, sangnom but never got a good resul (it doesn't seems to be ivtc only nor frame blending only).

The DVD is the USA version of "Otaku No Video" an old gainax anime form 1991.

Here's a sample of 20s of the vob source (m2v extracted with dgindex).

Sample.m2v (ftp://ftp2.yatoshi.com/yatoshi/sample.m2v)

Thanks a lot for helping

[mandarinka]

It is hard telecine, but in addition to that it has field blending (most dvd for old anime are like that).

You have two options, I'd say.
Option one: 1) use IVTC to match the fields; 2) since they were blended, you will have combed ghosts of past/future frames in the resulting video: 3) use some postprocessing to get rid of them. The ghosting itself isn't that offensive imho, in motion it isn't that visible (this might vary depending on the response time of your tv/lcd/projector though), it's mainly the combing that is annoying - IMHO.

To do this, I would use:


#Anti-aliasing with contra-sharpening by Didée
FUNCTION daa3(clip c) {
nn = c.nnedi3(field=-2)
dbl = mt_average(selecteven(nn),selectodd(nn),U=3,V=3)
dblD = mt_makediff(c,dbl,U=3,V=3)
shrpD = mt_makediff(dbl,dbl.removegrain((width(c)>1100) ? 20 : 11),U=3,V=3)
DD = shrpD.repair(dblD,13)
return dbl.mt_adddiff(DD,U=3,V=3) }

tfm(pp=0,mode=2,mmsco=false).tdecimate()
spline36resize(width,height*3/2).daa3().spline36resize(width,height) # blend-deinterlacing the residual combs with rather good detail-preservation.


Optionally, you could run mdegrain2 masked with tedgemask() after this, so that the blending gets a bit calmed on outlines.



Option two: 1) bob to 60fps 2) run srestore to get 24fps. This approach will lead to less residual blends, but is less detail-preserving.
It has certain bonuses - QTGMC tends to get rid of rainbowing and dotcrawl to some degree (but note that the above approach can still ofset that by using mdegrain2 on edges (tedgemask) for dotcrawl and other derainbowing filters - dfmderainbow with tedgemask and motion compensation tends to work well).

To do this:


qtgmc(preset="slower", MatchPreset="slower", MatchPreset2="slower", edimode="nnedi3",lossless=0,sourcematch=3,tr2=3,ChromaMotion=false).requestlinear() # chromamotion=false, since your source likely has rainbows, I assume?
srestore(frate=23.976)


P.S. The video doesn't look thatbad, so I would probably attempt approach one first, unless that results in too strong aliasing somewhere.

[Shinobu]

Thanks, the first script was not bad but some frame wasn't deinterlaced, so i added Tdeint() and it's all right, no more interleacing, blended frames remains but with a normal view speed on my lcd, it's not so bad ^^.