strange pattern to ivtc

[turbojet]

At points in this 30fps clip (http://www.sendspace.com/file/f7s9e8)there's a strange pattern after tfm/telecide, between frames 120-138 there's this:
D------D-D-D------D

D is dupe, - should be kept. The majority of it is a good 1 dupe in 5 frames. Is there anything that can remove these frames in a strange pattern while keeping the rest intact?

[wonkey_monkey]

Note to downloaders: sendspace apparently hopes you'll download their "accelerator," a .exe that does who-knows-what (servers adverts by the sound of things).

Untick the box before downloading!

(this direct link may work: http://fs03n2.sendspace.com/dl/ad21e4f2d5bd0d4a1c48ec526868576a/528166b550d9f601/f7s9e8/pattern.mpg)

[manono]

Part of the problem is that there are a few blended fields during the part you're having a problem with. Another is that there are missing fields. Maybe some were removed to make it play faster to keep it within the time allotted or to allow more time for more ads. I don't know. But the effective framerate is greater than 23.976fps. Anyway, following this script (first used for something else by jagabo over at videohelp.com), all your dupe frames are gone, but there are three missing frames. Even your basic IVTC script, while it produced dupe frames, also removed a number of unique frames.

Interleave(TFM(field=1), TFM(field=0))
Srestore(Frate=23.976)

[turbojet]

That's an interesting bobber which makes the missing frame less noticeable then tfm(mode=3).tdecimate(). Thanks, I'll keep it in mind if I run across the situation again.

For the whole encode I did a week ago I ended up using tfm(mode=3).tdecimate(hybrid=1) which made the missing frames less noticeable but added blended frames. In retrospect it was probably a bad thing to do.

[turbojet]

Any ideas how to ivtc this 60p clip (http://www.sendspace.com/file/cudsju)?

[ChiDragon]

Certainly can't be completely restored. The sample contains blended frames where the pattern breaks, ex: 152-174. The blended frames are also green-tinted (lacking red). Some frames appear to be missing (there is a jump between 550 and 551 for example).

[manono]

I agree. It's another case where frames have been removed and others blended, perhaps to have it play faster/shorter. This works reasonably well:

SRestore(Frate=24.3)

If you really wanted to take the time you could interpolate the missing frames before decimation.

[turbojet]

Thanks that is a slight improvement over tdecimate(). I also tried 2 pass vfr with tdecimate, results weren't much better.