PAL blended fields cartoons need help

[Great Dragon]

Hi guys!
There is nasty blanded fields cartoon video. Can't figure out how restore it. Restore24 and Mrestore are't restore it or maybe i'm not configured them properly.
Here is 2 samples: Sample1 (http://rapidshare.com/files/9512257/spider_sample.demuxed.m2v.html), Sample2 (http://rapidshare.com/files/9513497/spider_sample2.demuxed.m2v.html)
Thanks for your help.

[Great Dragon]

So, nobody can't figure out how to solve this?

[chipzoller]

I remember taking a look at this some time ago and couldn't find any good method to restore the footage. That's the problem with blended fields--sometimes it's just going to be impossible to undo the damage that was done in creating this mess. Fixing blended fields, at least to my observation, represents one of the biggest challenges to filter/application developers because it calls for an almost humanistic sense of intelligence when examining the frames.

Maybe some of the wise-men on the forum here (Didee, mug funky, neuron2) could help you better in this area than I.

[Great Dragon]

Maybe some of the wise-men on the forum here (Didee, mug funky, neuron2) could help you better in this area than I.
Hope so...

[Mug Funky]

it's a matter of whether there's clear fields in there. often the telecine process makes the blending last over several frames, especially on very bright objects. sort of like the lag that LCD monitors exhibit, but worse.

if the source is like that, then no amount of finessing will fix it completely. maybe some parts can be improved, but there'll always be frames blended together.

i haven't watched the sample btw - it might be okay.

Maybe some of the wise-men on the forum here (Didee, mug funky, neuron2) could help you better in this area than I.

i'm flattered, but i'm more in the business of causing the problem than solving it :)

[chipzoller]

Well maybe if you have a spare few moments you could take a look at his source? I'm interested in learning more about detecting the causes of blended fields and of course the various (imperfect) methods of trying to restore it.

[Mug Funky]

that video is awful. looks like ex-VCD. my suggested filter is Fire, and lots of it.

i'm guessing you've got a case of 3:2 NTSC being resized down (blending 2 out of 5 frames) to VCD at 29.97, then being run through a standards converter to get it to PAL.

now, there's a script out there to restore "double blended telecine", and several to restore "field blending" (actually should just be called standards conversion), but i'm not sure there's enough good info left in that to be able to do both.

as far as "understanding field blending" goes, the causes are these:

1. (most common) standards conversion. the source has been run through hardware to change the video standard (NTSC or PAL).

if you're lucky the machine used to do this will leave enough vertical detail in for a decent "un-blending". some are kinda ugly, and others are actually extremely advanced but very hard to undo (take an Alchemist PhC and run 3:2 through it and you'll end up with problems, but that doesn't mean the machine isn't very good...).

2. (not as common) old-school telecine machine. these will mix the tops and bottoms of different frames in some fields.

often they'll also have a sort of lag where bright objects persist several frames afterward. darks are usually very muddy and not much detail can be got out of them. depending on how clean the source is, you might be able to get a non-blended encode out of it. my suggestion is to just encode interlaced.

note that these machines will behave the same in PAL and NTSC, so acquiring the original format may not help you.

3. human error. someone's resized an interlaced picture, or done something else to muck up the interlaced structure.

this is uncommon but getting more common as the design and video areas get more similar - it's common practice to design for TV without actually having a TV to check it on. it's also sadly common that commercial packages don't expose field-order to the user properly (ostensibly to make it easier, but it really makes things worse in the long run - what the user doesn't know can in fact bugger up a project). this means the user is not aware of interlace at all, and it's up to the software to deal with it all properly - which it never does without user intervention, and thanks to the way these programs are set out the user never knows there's a problem until the disc is authored.

hehe.. probably more rant than information in that lot, but that's pretty much where it comes from and hopefully some insight into how to deal with it.

[MOmonster]

You won´t get any nice output with this source. As allready were said, it´s awful. There is not only doubleblending with different blendweights (not restorable with actual scripts) there are sometimes until six fields blended without any clear frame between (blends will be left so or so). With Cdeblend(omode=3,bthresh=0.2)+decimator most blends are cleared, but motion will be less fluid.
You should really look for an alternative source.:o

[chipzoller]

Thanks for the (semi)explanation as to the cause. Can both of you list plugins/filters used to treat this kind of material? I'd like to compile a list and collect any and all of these filters to try and experiment on different sources. I only know of 2 right now and those are restore24 and MOmonster's package (R_pack.zip) which contains Mrestore, Cdeblend, and FixBlendIVTC. What others are worth serious attention that have a better-than-normal chance of treating this type of material?
And just for clarification, how would you classify this type of blended fields material? Is this called a double blend? I noticed when stepping through with AssumeBFF().SeparateFields().Bob() that it looks like a patter of 2 fields of 1 prog. frame then 2 fields blended from the previous prog. frame + next prog. frame. Something like:

(fields)
1 - whole-------\ prog. frame
2 - whole (dup)-/
3 - blend of 2 and 5
4 - blend of 2 and 5
5 - whole-------\ prog. frame
6 - whole (dup)-/

It looks like the second of the blended fields in each sequence of 2 is either a stronger weight (leaning towards the next whole field) or alternating either top or bottom field of next frame (I don't know if this description makes any sense).
I have a source from a commercial DVD release that I was trying to examine yesterday that looks like it has the same pattern (http://www.megaupload.com/?d=Z1L520A0).
Is there any sure way of knowing what's caused this travesty?

[Mug Funky]

re-read momonster's post... the is no filter that stands a chance on this material.

even if you were able to restore clear frames, you'd still have to deal with everything else that sucks about it. this is not a typical source, and it doesn't make sense to waste time on a custom solution when it's still going to look awful in spite of it.