Battling Field-blended PAL material

[CoRoNe]

Hi,

I don't know everything about deinterlacing, but I always thought "TDeint()" would be oke for me (anime). It seems that if I compare two avs-files from the same video with AvsP the one with TDeint() even looks worse then without any deinterlacer at all :confused:...I must be doing something, don't I?
I think my video is Telecined, but I'm not sure about that.

Can anyone tell my how to setup TDeint properly for the following video material?
http://img148.imagevenue.com/loc1117/th_67472_sample_122_1117lo.jpg (http://img148.imagevenue.com/img.php?image=67472_sample_122_1117lo.jpg)

[foxyshadis]

Did you check the cadence by using SeparateFields()? That way you can also check that the chroma isn't broken, which would cause that sort of artifact. Otherwise, normally IVTC is what anime needs, though old ones might need Vinverse or something like TFM(clip2=TDeint()).

[CoRoNe]

Thanks for your answer foxyshadis, but honoustly this is beyond my knowledge.
I did manage however to take a look at the seperatefields like you mentioned. Even those "artifacts", like you call them, are on many individual fields.
Nevertheless, perhaps you can take a look at this (http://www.megaupload.com/?d=1V04PTK3) sample if you'd like. I think you have much more experience with such things than me.
I'm also not that familiar with all of TDeint it's parameters and what would fit my video most, so I would really appreciate it if you would post the commandline too :o
P.s. Is IVTC just another name for TDeint?

Thanks very much!

[Chainmax]

Look at the links in my signature ;).

[CoRoNe]

After some more heavily reading and testing in the early morning :rolleyes: I discovered I needed Decomb (Telecide) indeed. Man what a difference this makes! I always thought, deinterlacing...and that's it. :o
Thanks though, Chainmax.

[Chainmax]

Anytime :).

[CoRoNe]

After really A LOT of reading here and AvsP-testing with numerous avisynth-filters I realised Telecide() wasn't really the best way for me to get rit of all the ghosting effects. (the stuttering was really annoying!)
The demuxed.m2v (which is uploaded here (http://www.megaupload.com/?d=3C39NM2T)) is a sample of the Dragonball Intro from a Region4 (PAL) DVD which appears to be TFF Field-blended (like I said on a previous post; the "ghosting-effects" are even on separate fields, it's the source).

I discovered that of all the avs-filters I tried, Mrestore produced the best image (removed almost all of the ghosting-effects), but my intension has always been to keep the original framerate of 25fps...well let me just show you my avs-script:
mpeg2source("C:\test\DB - Intro(test).d2v")

TDeint(mode=1)
Mrestore(quality=2,numr=1001,denm=2002)

crop(6,2,-4,-2)
Lanczos4Resize(512,384)

LimitedSharpenFaster()
aWarpSharp(8.0,2,0.5)
Tweak(sat=1.05,cont=1.02)
Deen("a3d",1,3,5)
(of course DGDecode.dll, TDeint.dll, Mrestore.avsi, LimitedSharpenFaster.avsi, mt_masktools.dll, aWarpSharp.dll and Deen.dll are in the Avisynth plugins directory.)

I know Mrestore is originally intended for PAL > NTSC and NTSC > PAL restoration (right?), but numr=1001 and denm=2002 seems the give really good 25fps result (especially compared to Telecide, TFM and such).
The result: MKV[H.264] (http://www.megaupload.com/?d=U5YJEZLX)

Is this method recommended, or are there other options?

P.s can anyone tell me if there are any differences between "field-blended" and "field-shifted" material?
(oh and just to be clear this DVD is in my personal possession)

[neuron2]

P.s can anyone tell me if there are any differences between "field-blended" and "field-shifted" material? Yes, there is a difference. Field-shifted material does not have blended fields. But the fields that make up each frame are separated into different frames, because there is a one-field shift. Like this (top row is top fields, bottom is bottom fields, a letter is a unique picture):

. A B C D E
A B C D E .

[MOmonster]

Did you count the fields? Are there really 25 different clear frames in one second?
For 50->25 fps you can just use numr=1,denm=2.

I can´t download the source sample.

[CoRoNe]

Sorry MOmonster, I wasn't notified by mail by the forum for your post.

It really is PAL material and has 25 frames in one second.
Whould numr=1,denm=2 make a lot of difference compared to 1001,2002? (CPU power?)

My sample at rapidshare.com ( http://rapidshare.com/files/45618814/DB_intro_TDeint_mode1__Mrestore_Q2__.mkv.html)

[manono]

It really is PAL material and has 25 frames in one second
He's asking about the fields. Are there 25 clear, unblended, and unique fields in every 50? He's trying to have you determine if the framerate is supposed to be 25fps, or perhaps 24fps.

If you have trouble with the fields, even zoomed, then maybe put on a smart bobber, such as LeakKernelBob(Order=1), Yadif(Order=1,Mode=1), or TDeint(Mode=1).

[CoRoNe]

Even those "artifacts", like you call them, are on many individual fields.
Nevertheless, perhaps you can take a look at this (http://www.megaupload.com/?d=1V04PTK3) sample if you'd like...

So, blended fields then I guess. These episodes have most likely under went a NTSC --> PAL conversion, explaining the "artifacts".
But like I explained above, ultimately I'd still like to have 25fps.

[foxyshadis]

You'll first have to reverse it to 30 or 24 fps, whatever the original was, using MRestore. Then you pick your favorite method to convert it back to PAL in a nicer package, you'll probably be interested in the "closest results to alchemist" thread in that case.

[MOmonster]

Sorry MOmonster, I wasn't notified by mail by the forum for your post.

It really is PAL material and has 25 frames in one second.
Whould numr=1,denm=2 make a lot of difference compared to 1001,2002? (CPU power?)

My sample at rapidshare.com ( http://rapidshare.com/files/45618814/DB_intro_TDeint_mode1__Mrestore_Q2__.mkv.html)

I mean the source sample and not the output.
Foxyshadis said it! You have to restore the original framerate, also if the wanted framerate is another, else the restoring will not work.