Decomb 4.05 for Avisynth 2.5

[neuron2]

Merry Christmas!

Attached is Decomb 4.05 beta 2 for Avisynth 2.5. It supports both YUY2 and YV12 color spaces.

Please report bugs here.

[zymbalist]

A very Merry Christmas to you, dear Neuron2. I was really looking forward to decomb for avisynth 2.5....

[ArdenDag]

So this will worth with AVS 2.5 and 2.0x?

Also, did you discover the reason why decomb was failing on clips with low level field noise (I think that was what ya said... *grin*).. the clip wotef supplied you a while back. If ya did, is it fixed in this version, or are the agg and nt variables still configurable when need be?

Anyway, merry christmas :)

[neuron2]

Originally posted by ArdenDag
So this will work with AVS 2.5 and 2.0x?No. This version works only with Avisynth 2.5, but it supports both YUY2 and YV12 color spaces. Use DecombLegacy for Avisynth 2.0x.

Also, did you discover the reason why decomb was failing on clips with low level field noise (I think that was what ya said... *grin*).. the clip wotef supplied you a while back. If ya did, is it fixed in this version, or are the agg and nt variables still configurable when need be?Yes, it is all explained in the relevant thread. The nt and agg parameters are included with this new version.

[InfoCynic]

I've been using Decomb 4.03b6 for YV12 with AVIsynth 2.5 (071202 build). Getting back in the "encoding groove" after a gluttonous holiday break, and updated Decomb to 4.05b2 (modified 25/12/02) and AVISynth to 2.5 (171202 build). Renamed old versions of Decomb, and copied new decomb.dll into my plugin directory, making two copies, one as "decomb.dll" and one as "decombyv12.dll" so my old scripts wouldn't have to be rewritten.

AVISynth simply refuses to load the plugin. (C:\blah\decomb.dll is not a valid AviSynth 1.0 plugin). Version on AVIsynth reports 2.50 beta 1, date somewhere back in November. Guess the developers aren't updating that field or something. :)

Switched back to Decomb 4.03b6 and it's working for me, but if any of you have any ideas why the new one isn't, I'd love to hear it :)

[neuron2]

That's the error you get when you try to load a 2.5-style DLL using the old-style Avisynth. Try again, being very sure that the Avisynth invoked is the new one.

I have no problems running the new Decomb and the 12-17 Avisynth.

[neuron2]

Here is Decomb 4.06 beta 1 for Avisynth 2.5.

It fixes writing of a garbage character to the screen when show=true with guide=1.

I added a FILM/VIDEO indication when show=true and guide=1. You can easily see pattern breaks at scene changes, etc., because it changes from FILM to VIDEO and back again. Just a little something to justify another release. :)

[InfoCynic]

Tracked my problem to a rogue AVISynth DLL in the VDubMod folder. Loaded 4.05 and 4.06 no problem now. Thanx.

[neuron2]

Good news, InfoCynic. You caught a rogue. I like that.

@all

OK, OK, here's some source code:

[sh0dan]

@neuron:

Had a look at the source (actually I'm trying to do some mmx for it ;) - this was strange to me:

In fielddeinterlace/motiondetect, why is it val = ((long)p[x] - (long)srcp[x]) * ((long)n[x] - (long)srcp[x]); and not val = abs(((long)p[x] - (long)srcp[x]) * ((long)n[x] - (long)srcp[x]));
In my eyes, the threshold test would fail if one of the subtractions gives a negative result..?

[neuron2]

@sh0dan

You want the test to fail because that is not an instance of combing.

Suppose we have our vertical 3 pixels as:

p
c
n

To be combed, either c < p and c < n, or c > p and c > n. The metric gives a positive result only in those cases.

BTW, that code fragment corresponds identically to the code in Telecide that does postprocessing. That has MMX optimization started. I just haven't finished it and applied it to FieldDeinterlace yet.

[neuron2]

Here's another beta version. It fixes a serious bug in Decimate for YUY2.

It also implements an experimental new feature, hinting between Telecide and Decimate. I have initially used this feature as follows. When using Decimate(mode=1) the blends should be performed only on film sections of the video. So now, if you enable guide=1 Telecide will pass a flag with each frame marking it as from film. Decimate uses that flag to limit the blending only to film portions. This is intended to address a problem reported by StereoDude. Later I plan to use hinting in other ways (such as decimating combed frames, etc.)

A limitation of the hinting is that Decimate must immediately follow Telecide in the script.