Good Script lines for Decomb!

[datauser]

I once read the online readme for Handbrake and it stated that it is safe to leave the Decomb option on all the time whatever source. Yesterday I read the manual that goes with for this excellent dll tool and used settings to recover progressive frames with its plugin below to recover these frames rather than using the other hybrid settings given in the tutorial.

AssumeTFF()
Telecide()

I did a prior testing of the interlace pattern: analysis with another tool and my 23.976fps video showed up slightly as a hybrid with some frames being non progressive. My question is if by some chance the video is pure progressive and the analysis was wrong would the video become damaged?

Also what are good settings so Decomb can be left on so to speak for every input source whatever framerate? Thanks.

[Guest]

Originally Posted by datauser

I once read the online readme for Handbrake and it stated that it is safe to leave the Decomb option on all the time whatever source.

That is not true. Here are some ways it can go wrong:

* field matching of pure interlaced material can make a mess

* decimating telecined material is OK but decimating pure video will make a mess

* postprocessing deinterlacing can alter progressive material

That's just what I can think of off-hand. Throw in field-blended sources, field-shifted, etc. and the ways to go wrong increase in number.

My question is if by some chance the video is pure progressive and the analysis was wrong would the video become damaged?

For your example, probably not significantly. If it is progressive, the field matching will be OK. You may get minor effects from the postprocessing deinterlacing.

Also what are good settings so Decomb can be left on so to speak for every input source whatever framerate?

Don't go there. Just the decimation issue alone kills this idea. There just is no universal recipe for processing all videos.

[datauser]

Oh I see! Their FAQ gives bad info???, but down to subjective views. Anyway I take it from your meaning that this advice is quite wrong. Thankyou for your input.

https://trac.handbrake.fr/wiki/Decomb

Directly quoted from their link:

So what's decomb do differently?
The decomb filter looks at each pixel of each frame of a video. It then only deinterlaces frames that show visible amounts of combing.
This means you never have to check if a video you're encoding is interlaced -- just run the decomb filter all the time and it'll take care of everything.

[Guest]

Handbrake has appropriated my tool's name "Decomb" and apparently applies it to something like FieldDeinterlace(), which is a deinterlacer capable of deinterlacing only in areas with a spatial frequency close to combing. I answered relative to my tool Decomb, which also handles telecined material and hybrids. You also mentioned my tool.

Applying a deinterlacer that tries to detect combing can still cause degradation for progressive material because the combing detection is not perfect. And as I said if you apply it to telecined material you'll make a mess.

Stop looking for a universal recipe! It doesn't exist and you must tailor your processing to the nature of the video to get best results.

[manolito]

Yes, it can be a little confusing how Handbrake "invents" new names for things which already exist in the AviSynth world under a different name. Its Decomb function has absolutely nothing to do with neuron's Decomb package.

Having said this I must admit that Handbrake's Decomb function works exceptionally well. It uses quite sophisticated algorithms to detect combing. See the explanation in their Wiki:
https://trac.handbrake.fr/wiki/Decomb#function

Originally Posted by neuron2

Stop looking for a universal recipe! It doesn't exist and you must tailor your processing to the nature of the video to get best results.

That's true, but Handbrake's Decomb function does get pretty close IMO...


Cheers
manolito

[datauser]

Originally Posted by neuron2

Stop looking for a universal recipe! It doesn't exist and you must tailor your processing to the nature of the video to get best results.

Hey there, I'm not that naive to think this and by tailoring one's needs to the video to achieve "best results" sort of negates that as best results are always subjective are they not or even surmising what they might be before seeing?

And the Handbrake Team state it is also a modification of Yadif, but like manolito intimates here, it works rather well. I now see what the attributes of what neuron2 means in writing: appropriating my tool's name which forgive me as I did not realize this before in my following on in this line of inquiry. Thanks guys.