Cleaning up nasty interlacing

[linyx]

Hello all of you AviSynth geniuses,
I'm working on the first season of Lock N' Load (NTSC DVDs), but they are interlaced in some horrible manner that is very hard to remove.

My favorite attempt is:
MPEG2Source("Video.d2v")
SeparateFields().SelectEven().NNEDI2(field=1,dh=true)

But that leaves artifacts like this in the background which is very distracting:
http://i32.tinypic.com/2dv7a8g.png
Notice the "twisted" appearance on the pole.

I'd like to avoid double rate deinterlacing because these are already taking up a lot of size, but if there's something significantly better, then it's certainly worth it. Forgive this noob's lack of deinterlacing knowledge.:stupid:

Here is a decent sample (http://www.mediafire.com/download.php?1vkvijzv3j9x647).
I appreciate any advice you can offer,
Thanks in Advance!

[manono]

The best deinterlacer is one of the TempGaussMC variants. It's a double-rate deinterlacer, sure, but just add an SelectEven() afterwards. Your sample didn't include the curved and twisted goalposts, so I have no idea how it did with that. You do know, don't you, that if this is being reencoded for DVD there's no need to deinterlace at all? And that deinterlacing script of yours looks mighty peculiar to me. Did you just make that up?

[Didée]

Your source is not "interlaced in some horrible manner". It's simple, normal, standard interlacing. Your're just using a suboptimal way of deinterlacing. NNEDI2, as good as it is, is not a full-fledged deinteracer - it is "only" an interpolator, i.e. it is only guessing pixels within one frame. A full-fledged deinterlacer will also use data from the previous/next frame when possible, i.e. using temporal information.

I won't say anything against using TGMC.;) - However, did you already try the usual suspects, like TDeint, Yadif, YadifMod? Those are the most-used common deinterlacers, and do a complete job within their conceptual scope. NNEDI simply is not meant to be used as a standalone deinterlacer.

[linyx]

The best deinterlacer is one of the TempGaussMC variants. It's a double-rate deinterlacer, sure, but just add an SelectEven() afterwards.
I ruled TGMC out as being too slow, but WOW... I'm sold.:eek:

You do know, don't you, that if this is being reencoded for DVD there's no need to deinterlace at all?
Absolutely, but sadly it isn't.

And that deinterlacing script of yours looks mighty peculiar to me. Did you just make that up?
Mostly yes; I've (thankfully) only ever dealt with 3:2 pulldown, and never this stuff before, thus my lack of knowledge about it.

Your source is not "interlaced in some horrible manner". It's simple, normal, standard interlacing.
:stupid: But my sympathy stills goes out to whoever has to deal with this on a regular basis.

Your're just using a suboptimal way of deinterlacing. NNEDI2, as good as it is, is not a full-fledged deinteracer - it is "only" an interpolator, i.e. it is only guessing pixels within one frame.
I figured as much.

A full-fledged deinterlacer will also use data from the previous/next frame when possible, i.e. using temporal information.
I did not know this.

I won't say anything against using TGMC.;) - However, did you already try the usual suspects, like TDeint, Yadif, YadifMod? Those are the most-used common deinterlacers, and do a complete job within their conceptual scope. NNEDI simply is not meant to be used as a standalone deinterlacer.
Everyone has told you this at least once, I'm sure; but you are my hero!:D TGMC is amazing, even compared to Yadif and TDeint (which were better than my way by a fair bit). It's not that slow either, I was expecting much worse.
Thank you both so very much for helping this noob out!