GreedyHMA and avisynth 2.5

[sapient]

Hi...

Can greedyHMA be used with avisynth 2.5? Is there a special version of it? I am dealing with interlaced PAL content made from telecined NTSC and only greedyHMA seems to be able to straighten it out.

[trbarry]

Sorry, because of the internal data structures used there is no easy way to port GreedyHMA to YV12, though I supposed I could make a Avisynth 2.5 YUY2 only version if there was any real use for it.

But I think with all the fancy options now you will find Decomb has pretty much subsumed all the GreedyHMA inverse telecine functions as long as you take the time to figure out what to specify.

But these days almost everything can be handled either by Decomb or (for pure interlaced video) TomsMoComp(?,5,1).

- Tom

[Guest]

Thanks for the plug, Tom!

@sapient

If you can show us the issue you have, maybe we can help you with appropriate Decomb settings.

[sapient]

I am trying to encode some tv episodes (PAL DVD versions of Star Trek TNG and The X-Files.)
The original material is in NTSC, half of it true interlaced (the cg parts)and half telecined from 24fps. This source was transformed into interlaced PAL for europe using a truly stupid way, resulting, every once in a while, in PAL fields consisting of the blending of 2 NTSC fields , although the wrong two (that is not the identical 2, in the case of the telecined parts, but each of the identical two with the previous or next). Most of the rest of the frames are either telecined PAL or progressive PAL, varying from scene to scene, depending on whether the scene started in the first (top) or second (bottom) field of a particular frame. There are also some true interlaced parts (cg scenes). The result is an interlaced PAL stream that contains progressive parts, telecined parts, true interlaced parts and some pre-blended fields. This is what I am trying to straighten out.
Telecide seems to work, somewhat, but results in a really choppy image. GreedyHMA gives a much smoother image. I think that the mixed nature of the material confuses telecide which drops some frames or something. Plus, greedyHMA's postprocessing does a much better job of clearing up the fields already blended in the source; so good that, combined with 1-pass encoding with a quant of 2, the resulting divx actually looks better than the botched dvd.

[Fasola_Jan]

Hi :)

95 % of PAL X-Files material is field shifted every scene, so it is possible to reconstruct progressive frames perfectly (there is VirtualDub filter called “Deinterlacer – Auto PAL movie” made by famous Simon Walters, that can show You phase of field shift in given scene). I am sure that Decomb (by famous neuron2 – known to the rest of the world as Donald A. Graft) is able to do the same.

StarTrek TNG DVD PAL material is, as You said, field blended, so progressive frame reconstruction seems to be not possible. I am thinking about ST TNG PAL DVD conversion about a month now and nothing comes to my mind (even tried Photoshop picture substraction).

Good luck.

Bye.
Fasola

[sapient]

For the X-Files:

What you say is true, for the most part. The main problem is with the other 5% (or more) that is cg material created in true interlaced NTSC. This has been converted to PAL in a way that results in blended fields.

The same is true for the TNG cg scenes.

For the rest of the scenes, there is a blended field every 5 (if I remember correctly), but the rest can be reconstructed. Both decomb and greedyHMA can do that. The problem with decomb is that it gets confused by the blended fields resulting in a choppy image. Greedy HMA does not have this problem. Plus its postprocessing filter does a marvelous job of clearing up the pre-blended fields. I have used greedyHMA((1,0,1,0,1,0,0,0) for TNG (PAL) with great results.

So, as long as decomb continues to have this problem, a version of greedyHMA for avisynth 2.5 would be great. Until then I will stick with version 2.0x... unless, that is, any of you have any other suggestions...

[sapient]

After all these months, I still find myself depending on GreedyHMA to de-interlage PAL material derived from mixed video and telecined NTSC, as is the case for most TV-series and the making-of and featurets of movie dvds. It is without doubt the best for this kind of thing. Telecide still and consistently makes a mess of these clips, no matter how much I try to tweak its parameters... So ...

Tom, could you, PLEAAAASE, make an avisynth 2.5 version, even if it only supports YuY2?

sapient

[trbarry]

Tom, could you, PLEAAAASE, make an avisynth 2.5 version, even if it only supports YuY2?

I use Decomb for most things myself these days and DG is probably the best available expert now for handling pathological cases of telecine/video mixes. So you might ask him about the problem cases.

But I will create and post a YUY2 only version of GreedyHMA for the folks that still want it. I'll post here when it's out.

- Tom

[trbarry]

Okay, I made one. Version 0.4.1 is compiled for Avisynth 2.5, but still only supports YUY2. It's now on the downloads section of my web page.

There's also still an older Avisynth 2.0 version in the zip.

- Tom

[sapient]

That was fast!
Seems to be working fine.

Thanx!

[jarthel]

any chance of a yv12 version? ;) it seems to me that for same sources greedyhma is still the best. I verified this with an anime awhile back.

[trbarry]

any chance of a yv12 version?

jarthel -

Sorry, probably not. The internal data structures ported from an older version of DScaler make it harder to do.

But mostly, any IVTC function has to deal will a million special cases and options of pathological telecine. Neuron2 seems to have the skill & patience to add options to deal with these, but I don't.

So if you have specific examples of clips where GreedyHMA outperforms I would probably pester DG about it. He has been pretty good at adding any code where mine does something better. And for the most part the Decomb package has gone far beyond my quick & dirty port of GreedyHMA. I only made the Avisynth version because I didn't believe there were any decent Avisynth IVTC packages at that time.

I really haven't made any attempt to improve GreedyHMA for over a year now except to compile & test it for A2.5 this week.

- Tom

[sapient]

I would love to pester DG about it but,the trouble is, I have no idea how to contact him. As far as I know he no longer reads this forum and his web page has gone down.

[bilu]

http://www.neuron2.net

http://neuron2.net/ipw-web/bulletin/bb/index.php


Bilu