tomsMoComp as a postprocessor for telecede/pfr?

[sapient]

I have been experimenting again with various combinations, trying to get the best result for deinterlacing those bad pal transfers. I seem to get pretty decent results using TomsMoComp after pfr, using it, I hope, the same way telecide uses field-deintrlace.
The readme for TomsMoComp states though that it is intended for pure video material and I was wondering, if there is some side-effect to what I am doing that I haven't thought of.
Anyone?

sapient

[wotef]

i don't think doing this gets you anything different from just running tomsmocomp alone

[sapient]

Hopefully pfr reorders any fields that can be reorder, producing some progressive frames that do not need to be de-interlaced by TomsMoComp afterwards, like telecide does before running field-deinterlace as a post-processor.

[trbarry]

The readme for TomsMoComp states though that it is intended for pure video material and I was wondering, if there is some side-effect to what I am doing that I haven't thought of.
Anyone?

TomsMoComp can be applied to most anything to take out weave artifacts but on telecined progressive material it will not let through as much detail as Telecide, UnComb, etc.

I should probably add some preprocessing code like in Telecide or GreedyHMA to check if the frames are already just fine and leave them alone when possible. But I probably will not be doing that soon.

- Tom

[tempetai]

Originally posted by trbarry
I should probably add some preprocessing code like in Telecide or GreedyHMA to check if the frames are already just fine and leave them alone when possible. But I probably will not be doing that soon.

- Tom

Great, Just let us know once you have it ready. I can wait... TomsMoComp is a great deinterlacer...

[sapient]

@trbarry

Well, uncomb or telecide are not options here, as I am dealing with bad pal clips that confound telecide and are not suited for the simpler uncomb. I use greedyHMA, but because of the need to convert into YUY2, I 've also been looking into pfr. Pfr is smarter than uncomb, but like it, does not include any post - deinterlacing. What I am looking for is a de-interlacer to serve as a post-processor that will let through the maximum amount of details.

The options I have explored till now are, in order of smartness:
-TomsMoCOmb, who according to you, Tom, de-interlaces everything, even when they do not need it.
-FieldDeinterlace(), that de-interlaces only frames detected as combed, but once it has marked them so, de-interlaces the whole frame
-KernelDeint, that only de-interlaces the areas it detects to contain motion (and thus combing).

One would think that KernelDeint is the choise of prefence then, but actually it produces some weird artifacts in the de-interlaced areas.

So, other thoughts?

sapient

[bilu]

@sapient

Have you tried KernelDeint(sharp=false)?

I also got a lot of artifacts, but only with sharp=true (default).


Bilu