How should this be converted to progressive? (another 35mm capture)

[mantis2k]

What type of source is the capture? It doesn't appear to be Field-Blended this time around. My Phase/Field-shifted solution still left combs. Leakkerident doesn't seem to work, either. Is there permanent combs in the capture? (unlikely). Could somebody please analyse the fields and let me know how I should convert to progressive using Avisynth? I am not very technical when it comes to this. Any help would be much appreciated!
Download clip... (http://www.rarekungfumovies.com/source.mpg)

[Guest]

It doesn't appear to be Field-Blended this time around. It *is* field-blended. Why do you say it doesn't appear to be? Do you have a reliable way to tell or are you just guessing?

[mantis2k]

It *is* field-blended. Why do you say it doesn't appear to be? Do you have a reliable way to tell or are you just guessing?
Damn... is that really so? I already tried SRestore, but it didn't even come close to restoring that. :confused:

[Adub]

I agree with neuron2, there is blending all over the place in this clip. Just look at the fighters' arms or head as they duck and weave.

Edit: Frankly, srestore may also be getting confused by the junk on the left handside. Crop that and try srestore on it. All in all though, don't expect miracles on this clip.

[mantis2k]

Thanks. Could somebody please suggest how I should go about "trying to find a pattern" and "decimating"? I am a complete novice, but heard that's the best way of dealing with Field-Blended.

[Adub]

When finding a pattern, it is best to examine the source footage frame by frame. If you see a blend, count it, and so on. When you are done, you can decimate using filters like Decimate() and your pattern. I suggest you read this: http://server4.doom9.org/index.html?/decomb.htm

[mantis2k]

Thanks for your reply. Please can you (or anyone else) provide any links to more in-depth guides/articles on how to spot "blends", the correct way to note them down, and how to apply the analysis to the Decimate() function? Again, I am complete novice.

[bizz & buzz]

It's looks like the usual field-blended interlaced Pal source. It's not easy to find a solid pattern on this kind of source, as far as I know.

For me, Srestore gives decent results. Sure there are a few blends left here and there, but it's much faster then removing the blends manually.

MPEG2Source("D:\source.d2v", cpu=0)
Crop(62, 0, -0, -0)
tdeint(mode=1, edeint=nnedi2(field=-2), full=false, tryweave=true)
srestore()
nnedi2_rpow2(qual=3,rfactor=2,cshift="spline36resize",fwidth=1356,fheight=576)# funny
scaling, just in order to correct aspect ratio while maintain full vertical resolution.

[mantis2k]

Cheers. I'll give that script a try when I get home. I would still be interested to learn the manual option as well though...