Decomb weirdness (Duck Dodgers)

[litz]

Anyone have ANY idea how to properly decomb this?

I've tried about every parameter tweakable, and I either get weird shadows or incorrect interlaced results.

Using avisynth 2.5x and Decomb5

Any suggestions using the above or other filters to try, please followup back so others can gain.

Sample (huffyuv) is at :

http://terminus.litz.org/dodgerstest.avi

thanks !

- litz

[stickboy]

I tried Telecide(1).Decimate(), and AFAICT it gets rid of all the combing...

The "shadows" are from the source material; there's nothing you can do about them. (You can see this for yourself by calling SeparateFields() and observing frames 37-40.)

[litz]

I noted there's now a avisynth2.5 version of greedyhma.

So I tossed it at it.

Not needing any of the IVTC options, I set it simply as :

Greedyhma(1,0,0,0,0,0,0,90,115)

And got this result :

http://terminus.litz.org/dodgersgreedy.avi

Anyone know what the funny fuzzy stuff is?

Is there a better solution?

Tried the Telecide/Deimate described above. Doesn't work at all.

- litz

[litz]

For those interested ...

Ran the original clip :

http://terminus.litz.org/dodgerstest.avi

which is a segment of a straight huffyuv capture

running it through greedyhma yields :

http://terminus.litz.org/dodgersgreedy.avi

moving onwards, cropping and reducing to SVCD size (with a convolution3d filter added in) :

http://terminus.litz.org/dodgerssvcd.avi

and adding msmoother.dll in addition to the above yields :

http://terminus.litz.org/dodgerssvcd2.avi

Seems to clean things up rather nicely.

Anyone got a better way of going about it?

This way works ... but it's godawful slow ...

- litz

[Si]

Decomb (decomb500b16.dll) seems to work OK at my end as well.

Maybe you have an older/different version :confused:

regards

Simon

[litz]

It works *okay*.

But I still see quite a few blended frames and those shadows.

Remember, I have about 23 more minutes of video to work with - what you see above is just representative of the problem.

Running through GreedyHMA seems to remove all blended frames and the shadows go away too (I believe they're getting interpolated out).

The end result is quite nice. It's just SLOW.

- litz

[stickboy]

The fuzzy lines/shadows have nothing to do with the IVTC. As I said, if you call SeparateFields() and look at the individual fields, you can see that the fuzzy lines are already there.

They're either from the original source footage, introduced by compression on the cable operator's end, compression on your end when you captured it, etc..

[Guest]

stickboy is quite right. That noise is called mosquito noise and it comes from the compression process. You can adequately remove it using any decent 2D smart smoother. For example, use Decomb as described and serve it into VirtualDub. Then add my Smart Smoother filter and set radius=11 and threshold=70. Other smoothers may do even better (Smart Smoother was a very early effort in the evolution of our smoothing technology).

[litz]

Neuron,

Yah, I found out smoothing fixes that.

used "msmooth.dll" which seems to be optimized for animation and
the results are fantastic.

I still don't get why decomb misses so much interlacing, but since greedyhma seems to work pretty darned well, there IS a solution for this.

Interlaced NTSC is a major pain-in-the-***.

- litz

[Si]

I cannot see any "interlacing" artifacts on your test clip after applying Telecide(1).Decimate()

I can see lots of "mosquito" noise around the mouth in particular.

regards
Simon

[Guest]

Originally posted by litz
I still don't get why decomb misses so much interlacingI too did not see a single combed frame on your test clip. Care to elaborate on your complaint?

[litz]

I'll see if I can reproduce the combing I saw ...

Maybe it was something to do w/my particular setup, don't know ...

But I *did* see combing that decomb missed, that weird shadow effect (which is *gone*, when greedyhma is applied), as well as the aformentioned mosquito noise around the mouth.

Any ideas as to why decomb leaves the shadows and greedyhma does not?

someone else mentioned something about it being in the fields, but why would there be no trace w/greedy ?

Look carefully as porky falls out of the chair, then look again at the greedyhma clip : no shadow at all. And it doesn't seem like it is interpolating. It's a fairly sharp frame.

I think what this shows, in general, is that Cartoon Network and Direct TV need to learn how to broadcast video.

- litz

[litz]

Hmmmmmmpf.

Well, now I can't get decomb/avisynth to reproduce the combing effects I was seeing. I get combfree frames w/shadows, as described before in the thread.

So no idea whatsoever what was going on there.

I *knew* I should have run off another test sample.

:-(

Back to part 2 of my question before, though ...

Using either greedy or decomb to decomb this ... Decomb produces frames with those shadows and greedy does not.

Is greedy automagically interpolating? I'm not very up on exactly how greedyhma actually works.

- litz

[Si]

I think with your GreedyHMA settings you are asking it to do a straightforward video deinterlace of your clip.

With deinterlacing you normally (always probably) lose vertical resolution as some sort of blending of the fields is applied.

This is no doubt reducing some of your "shadows".

regards
Simon