Pal 25i Wide to Ntsc problem

[spiritgumm]

have a pal widescreen dvd which I'd like to slow down to ntsc with Avisynth. It appears to be interlaced, though I dont see any backward motion in either field order. I assume there's still no way to avoid deinterlacing to slow it down, so I thought I'd try out the latest QTGMC(), followed by Selecteven and AssumeFPS (23.976).
The jaggies diminish as you use slower speed presets in QTGMC. However, I found that "Fast" and slower made the edge between top of film and black border dance. So I'm not sure how to proceed. I suspect the jaggies in "Very fast" and "Ultra fast" might be equivalent to Yadif jaggies, and therefore not advantageous.
Sample of video:pal wide interlaced.m2v

[poisondeathray]

it's progressive content, but field shifted

there is significant aliasing, but I wouldn't deinterlace this (or as a last resort, because deinterlacing it will degrade the image , making it blurry). I would use QTGMC's progressive mode combined with vinverse

MPEG2Source()
AssumeTFF
SeparateFields
Trim(1,0)
Weave
Vinverse
QTGMC(inputtype=1, sharpness=0.5)
AssumeFPS(24000,1001)
Spline36Resize(720,480)

[poisondeathray]

Here are some comparisons of the methods, one is treating it as progressive (as above), the other is using QTGMC to deinterlace it and selecteven(), the method you initially proposed.

I included xvid encodes and some screenshots. If you examine the fine details on the stills (e.g. look at the chair, the man's face, the stone vase - fine details have been blurred away by the deinterlacing, but are preserved by treating it progressive). Notice the aliasing on the bus and minivan on the video. Now you could use AA filters or try QTGMC in progressive mode to process it again, but that will reduce the fine details even more.

[spiritgumm]

this is a response to your first post, not the 2nd.

thanks, Poison. So it looks like you're trying to minimize the combing and shimmer.
Was the Separatefields-Trim-Weave intended for the whole movie, or just this clip? If the movie, can you explain what it's doing?

Btw, I'm still using the same old slow PC. Would using Fast or Very Fast preset in QTGMC compromise the desired quality?

[poisondeathray]

Yes you would use it on the whole movie, assuming it follows the same pattern

It just looks interlaced, but it isn't . It's called "field shifted" or "phase shifted" . It's when the fields are misaligned

Normal progressive field alignment should be like this (field pairs are together and happy):
[a a] [b b] [c c] [d d]

Phase shifted: here "z" is in the wrong place, and the a a, b b, etc.. are not grouped together. This causes what you see as "combing" . So it looks "interlaced".
[z a] [a b] [b c] [c d]

So the you're trimming off the "z" field, and realigning them back to
[a a] [b b] [c c] ....


Another way you could do this is use TFM()

Try it out without any of the filters, just separatefields, trim, weave and preview it. You will notice no combing. But there still is aliasing, that's what the other filters are for

Btw, I'm still using the same old slow PC. Would using Fast or Very Fast preset in QTGMC compromise the desired quality?

a little bit, try the "faster" preset for the QTGMC progressive mode. The results are still better than deinterlacing it.

[spiritgumm]

The sample look good on the PC. I'll check them on TV.
I only mentioned deinterlacing because everything about it looked interlaced, even though the fields were progressive.
I tried TFM() when you said it was phase shifted, but it doesnt do quite as well as your trim-weave. So since it's progressive, there was no doubt you just needed to trim one field?

[poisondeathray]

I tried TFM() when you said it was phase shifted, but it doesnt do quite as well as your trim-weave. So since it's progressive, there was no doubt you just needed to trim one field?

There might be some other combinations, but the phase shift is the first one that comes to mind

If that sample is representative of the big movie it should work