New Cartoon Encode Problem

[IceManTX]

Here is a small snippet of source material:

obanbefore.mpg (http://71.164.234.15:81/moviez/obanbefore.mpg)


Here is a small snippet after converting to xvid:

obanafter.avi (http://71.164.234.15:81/moviez/obanafter.avi)


Here is the stuff I'm doing:

TFM(d2v=Myd2v,pp=5,clip2=TomsMoComp(-1,5,1))
tdecimate(mode=1,hybrid=1)
crop(10,4,700,472)
LanczosResize(640,480)
frfun3b(4,3,6)
vmToon(sharpen=false,thinning=0)
LimitedSharpenFaster(Smode=3,strength=80,overshoot=1)
warpsharp()


I am getting very wierd blending/interlacing/blur every so often that I cannot remedy no matter what settings I try and tweak.

Thanks for any help.

[IceManTX]

Has anyone had a chance to look at this?

I was up til 4am this morning adjusting and trying different things and I just can't figure out what I am doing wrong.

Any help is greatly appreciated.

[Pookie]

Tons O' Blends....:(

Look around for Cdeblend, Restore24, etc. Or search the term Anime Blend .

Perhaps the Fieldmatch Function ?

More:
http://forum.doom9.org/showthread.php?s=&threadid=66093



This looks quite improved, but there are still are a few blends.
http://www.megaupload.com/?d=G1XQTEII

MPEG2Source("E:\obanbefore.d2v",cpu=4)
converttoyv12()

Interp = SeparateFields().SelectEven().EEDI2(field=1)
deinted=TDeint(edeint=Interp)
TFM(mode=6,order=1,PP=7,slow=2,mChroma=false,Clip2=Deinted)
tdecimate()
fieldmatch()
crop(8,8,-8,-8)

backward_vec2 = last.MVAnalyse(isb = true, delta = 2, pel = 2, overlap=4, sharp=1, idx = 1)
backward_vec1 = last.MVAnalyse(isb = true, delta = 1, pel = 2, overlap=4, sharp=1, idx = 1)
forward_vec1 = last.MVAnalyse(isb = false, delta = 1, pel = 2, overlap=4, sharp=1, idx = 1)
forward_vec2 = last.MVAnalyse(isb = false, delta = 2, pel = 2, overlap=4, sharp=1, idx = 1)
last.MVDegrain2(backward_vec1,forward_vec1,backward_vec2,forward_vec2,thSAD=800,idx=1)
limitedsharpen(strength=320,Smode=4,ss_y=1.8,ss_x=1.8)

You can use a denoiser other than Mvdegrain2, unless you have the time (or a Cray). ss_y and ss_x helps reduce aliasing.

[IceManTX]

Thanks Pookie!

Now I have somewhere to start. This is a new animal for me. Everytime I think I have this encoding world down and I can relax, something else comes along to bite me in the arse.

[Awatef]

Actually, I would rather forget about converting this to whatever non-interlaced format if I was you. You won't get optimal results no matter what you do.
You better backup this to DVD or SVCD (keep it interlaced).

[MOmonster]

Actually, I would rather forget about converting this to whatever non-interlaced format if I was you. You won't get optimal results no matter what you do.
You better backup this to DVD or SVCD (keep it interlaced).
In general I would say you are right, this source is to blame, but in this special case, no.
Thanks Pookie!

Now I have somewhere to start. This is a new animal for me. Everytime I think I have this encoding world down and I can relax, something else comes along to bite me in the arse.
This fields that looks like blended are the really bad result of denoising without regard to the 2 of 5 fieldshifted frames. Indeed a simle blending looks much better for me. The only luck you have is that the temporal effect of the denoiser is so little (maybe no temporal effect) and the spatial effect is a simple (area)blurring.
To get rid of these strange artefacts you have to create real blends. This new wrong ivtc can be restored. Some little artefacts will be still there but the source looks much better. Here is a good script to begin with:
telecide(order=1, blend=true, vthresh=30, dthresh=3)

FixBlendIVTC(post=5) #sbd=true ?
decimate(cycle=5,quality=0)
It is not so easy to find out this trick, especially if this is a new animal to you.;)
I don´t know if your source is a 24fps animation or an anime with some 12fps scenes inside. For the second case set sbd=true.
For more informations read foxyshadis double-blend ivtc removal (http://forum.doom9.org/showthread.php?t=101477&highlight=ivtc+removal) thread and keep in mind that FixBlendIVTC is still in development.