Deinterlacers without field parameters

[nixo]

Hi

A friend of mine needs to convert a series of clips from different sources for display on his web page via flash. He has zero experience with video encoding so I'm writing a couple of batch scripts for him to use.

Some sources wil be TFF and others will be BFF or progressive. To make it easier, I'd like to use a deinterlacer that doesn't care about field order and preferably doesn't do too much damage to progressive material.

I've thought about using fielddeinterlace(full=false). I'm just wondering if there are other good choices?

--
Nikolaj

[LoRd_MuldeR]

If you want to completely ignore filed order, this means you will have to throw away one filed and interpolate the lines from the remaining one.

As soon as the deinterlacer looks at both fields, their temporal order must be known. So your options are very limited here. Bobbers won't be an option for sure.

Maybe FFPP("ffmpegdeint") or FFPP("cubicipoldeint") from FFVideoSource2 package would do the job...

[Didée]

As soon as the deinterlacer looks at both fields, their temporal order must be known.
That's not a must. Preferable, sure, but not a must. Deinterlacers looking at only +-1 frame are not seriously disturbed by wrong field order. E.g. Yadif looks pretty much the same, no matter if feeded with correct or wrong fieldorder. IIRC TDeint is also relatively robust in this respect. (The result is phase-shifted by one field, of course, but that doesn't really matter, usually.)
For bob deinterlacing it's a different story, there it is absolutly important. But for same-rate deinterlacing, much less so.

Pure NNEDI2 deinterlacing could be another failsafe option for the task, if time allows.

And, really, if the task is to provide a nobrainer script for someone who is not into these things at all, then just go for simple solutions. Discard one field & resize, tell him from heart that it's a brilliant script, and he will agree that the result is brilliant.

[nixo]

Discard one field & resize, tell him from heart that it's a brilliant script, and he will agree that the result is brilliant.

Heh. Yeah, I think I'll do just that. My thanks to both of you.

--
Nikolaj

[manolito]

Another option (probably faster than using EEDI2 or NNEDI2 for resizing) would be to use the old Decomb package v. 4.10 which is still available on Neuron's website. This version does not require a field order parameter.

Cheers
manolito

[neuron2]

That's for the field matching filter Telecide(). Even in the latest packages FieldDeinterlace() does not require field order.

[manolito]

Oops, sorry, I mixed it all up somehow...