Fieldblended content detection in AviSynth script?

[jeffy]

If there is such a topic, please just point me to it, thank you.

I would like to find the AviSynth script that is able after using "SeparateFields()" on a MPEG-2 TV stream decoded via mpeg2source to show me, that the content is fieldblended.

[jeffy]

I would like to extend my request also for the field shifted material. Can anyone help me please? If it is not possible within the AVS scripts, could you please give me a good piece of advice how to recognize it reliably? :thanks:

[Leak]

I would like to extend my request also for the field shifted material. Can anyone help me please? If it is not possible within the AVS scripts, could you please give me a good piece of advice how to recognize it reliably? :thanks:
Errr... the obvious solution to field shifted material is a
AssumeFrameBased().SeparateFields().Trim(1,0).Weave()
(off the top of my head, might need some adjustments) followed by watching a bit of the result. If the combing is gone, the material is field (nee phase) shifted.

As for field blending, again, SeparateFields() followed by *watching* the result is key.

Or were you hoping for an automated solution and just forgot to say so? :confused:

(But given that running an AviSynth script usually results in a video watching is an integral part, don't you think?)

np: Funckarma - Strip (Elaztiq Bourbon 5)

[jeffy]

Or were you hoping for an automated solution and just forgot to say so? :confused:



Thank you very much, yes, I was hoping that some AviSynth script could just tell me - this is with a 90% probability field-shifted material....

Something similar to MeGUI Source type detection, eg. Partially interlaced, do you know what I mean? Or is it just a dream :) and it cannot be done by a script?

Please if you or someone else can clarify, it would be great.

BTW, what is a nee phase, please?

[Mug Funky]

consider how long it takes a human to learn how to spot field blending... then apply that to a computer which is infinitely dumber.

it's possible, but it wont be 100% reliable (probably pushing your luck to get a correct guess even half of the time), and it'll be much slower than simply watching the footage.

you might be able to semi-automate it though - as in adjust your workflow so that the watching-it-to-identify-it stage happens painlessly (like a batch process that runs a simple bob with mplayer over a section, then prompts the user to enter "fieldblended? (y/n):".

it's hard enough making an automated 3:2 vs 30p vs 60i detection...

[Leak]

Thank you very much, yes, I was hoping that some AviSynth script could just tell me - this is with a 90% probability field-shifted material....
Well, you should be able to get away with just adding a call to TFM() (from tritical's TIVTC package, probably with some slightly tweaked parameters) at the top of your script - if you don't use TDecimate afterwards it should correct field-shifted material, but leave "real" progressiv frames alone.

It won't recognize fieldblended material, though.

[IanB]

Throw a set of candidate solutions into a testing script and use some StackVertical/StackHorizontal or Interleave to preview the candidate solutions in parallel. Choose the 1 that works best and run with it.

Make the computer do the hack work, let the human make the choice.

[jeffy]

Thank you to all. IanB, I will certainly try your suggestion!