Pixellation/stepping on video footage

[Morte66]

I'm trying to clean up some convert video footage from interlaced PAL DVD. It has a sort of pixellation effect that turns into stair stepping on diagonal edges, it looks a bit like there's been a point resize at some point in the production chain. Or maybe I did something terribly wrong in the deinterlacing. Here's a sample (http://www.mediafire.com/?bhmigidsneg). Have a look at the neck on the bass guitar, for example.

So far I've tried...SetMTMode(5)
DGDecode_mpeg2source("C:\MD\@MeGUI\Heima 206\Heima206.d2v",idct=4,info=3)
SetMTMode(2)

TDeint()
#vinverse()

Tweak(bright=-16, cont=1.164, coring=false)

DeBlock_QED_mt2( quant1=20, quant2=24, aOff1=2, bOff1=4, aOff2=4, bOff2=8 )

mvdegrain3p (source=last,blksize=8,overlap=4,sharp=1,thSAD=400,idx=1)

gradfunkmirror(1.2)

I suspect I'm not making the best use of deblock_qed_mt2 on this (de)interlaced footage. Apart from that, is there anything I can do about the pixellation?

[Didée]

The content isn't truly interlaced. It it progressive with dynamic phase shift.
The original recording most probably was interlaced, but has been *very poorly* deinterlaced at some point before or during DVD mastering.

Because of the dynamic field shift, normally you would start out with a fieldmatcher to get progressive frames again. However, in this case there's not much to gain from field matching: the poor deinterlacing process did also deinterlace (i.e.: interpolate) in static (!) areas, so that fieldmatching will only recover the interpolation.

Basically you have to throw one field out completely, and use e.g. NNEDI to interpolate full frames again. (Eventually you still have to do fieldmatching before cancelling fields, because of the dynamic phase shift. I didn't try, just have looked at the raw footage.)

[Morte66]

Wow.

Well, I think it's probably better to just admit defeat on this one.

Thanks for looking at it.