IVTC/decimation problem

[^Rogue^]

Hi everyone,

I've been trying to rip this to XviD but I can't get the framerate right it seems. I've played around with tDeint and mRestore. I can lose the combing but some lines at the top appear 'shifted' (don't know how else to decribe it).

elguaxo took a look and assumes it's some kind of "VHS to DVD atrocity".

Here's the sample (~30 mb):

http://rapidshare.com/files/113439804/Zeiram2_Sample.demuxed.m2v.html

Thanks in advance for you effort. :)

[45tripp]

how about a call to vinverse() after tdeint(1).
and then some smoothing.

it's pretty awful to begin with, but this should be passable:
http://www.mediafire.com/?jfznn9kmcjm

[^Rogue^]

Sorry for the late reply.

Looks better than what I have been able to achieve so far, thanks. Will look into it. :)

[thetoof]

You've got blended fields.
Try
tdeint(1) #or any other bobber... you could also add nnedi or tmm in there
selectevery(5,0,2)

[elguaxo]

Great! It works perfect for this scene, but I always have bad luck with this solutions (selectevery), because the same pattern is not kept for the whole DVD. If thats the case on this DVD, any hints how to help MRestore in this particular case to do the same job? This fails to do the same compared to the selectevery line:

#bobber and
Mrestore(numr=2002, denm=5005)

[thetoof]

The thing is, it's similar to the problem I encountered in the "Cartoon IVTC" thread. After bobbing, you have this sequence:
good frame // frame with Undetectable Blend // good frame // frame with UB // ok frame, but it's a duplicate of the next, which looks better
And well... UB says it... it's undetectable, so no I don't know any way to make it automatic.

[Didée]

BTW, what makes a blend "undetectable"? I'm not sure ATM if the actual MRestore still uses the same blend detection technique as the early versions did ... but the early versions used a very similar technique as Restore24 does, and the *raw* blend-detection of Restore24 at times surprised me with detecting blends that I could hardly identify even on very close inspection. (I'd even say that as long as there's only one single blend (not two ore more in a row), detection of blends isn't such a big deal.)

Perhaps there's need of more tuning of MRestore's parameters? See, if a function offers several parameters, but is called just with empty parantheses, then it's not necessarily fair to conclude "it doesn't work" ... ;)

[45tripp]

sorry,
noticed i didn't mention mrestore, when it was used to create the sample i uploaded.

what i actually used was:
tdeint(1)
vinverse()
mrestore(numr=40,denm=100)
STMedianFilter(8,15,4,7)
msmooth(strength=1)

[^Rogue^]

Thanks to everyone for the suggestions. :thanks:

Right now my work schedule leaves me with very little spare time for encoding but I'll try the solutions on the complete movie and get back to you as soon as I get some air. :)

[thetoof]

My bad! The blends were indeed detectable for this source and mrestore works. Don't know why it didn't when I tested it yesterday... I most likely had made some kind of mistake somewhere.
It also works on what I had called UB in "double 24 fps anime" (Cartoon IVTC thread), but only since I've been using tdeint(1,edeint=nnedi(-2),emask=tmm(1))
Only a few still remain after mrestore... maybe it could be removed after tuning, but I didn't find any good combination of settings to do so.
mrestore (http://img55.imageshack.us/img55/1601/mrestorega1.jpg)//Selectevery (http://img522.imageshack.us/img522/9528/selecteverycs8.jpg)
Have fun with this (http://www.mediafire.com/?bs4nwtzgzbj) sample.

[^Rogue^]

My bad! The blends were indeed detectable for this source and mrestore works.

Can you give me the parameters you used? :thanks:

[thetoof]

Actually, I used your sample at some point when I was writing AnimeIVTC() and the following gave me the best results I've seen so far (only for the ivtc portion of the script) on your source.
animeivtc(2, aa=0, killcomb=2)

[^Rogue^]

Cheers, will give it a go. :thanks:

[Blue_MiSfit]

This new AnimeIVTC script seems to be owning anime left and right. I can't wait to give it a try!

~MiSfit

[neuron2]

This new AnimeIVTC script seems to be owning anime left and right. What makes you say that, given the paucity of results reported thus far?

[Blue_MiSfit]

I get a vocab lesson today! Thanks Neuron2 :)

from m-w.com:

Main Entry:
pau·ci·ty Listen to the pronunciation of paucity
Pronunciation:
\ˈpȯ-sə-tē\
Function:
noun
Etymology:
Middle English paucite, from Latin paucitat-, paucitas, from paucus little — more at few
Date:
15th century

1 : smallness of number : fewness
2 : smallness of quantity : dearth


I'll be the first to admit, I tend to get overly excited about new things.

You're right in saying there aren't many results reported yet, but so far it looks to be a useful part of our arsenal. I'm also very excited by the prospect of a plugin that unifies some of the trickier things in this department like VFR. Finally, it seems to have comprehensive, well written documentation.

~MiSfit