Deinterlace methods

[bolix999]

Not sure if this should be here or in Capturing.

I'm having a running discussion with a colleague at work about capturing, deinterlacing and resizing.

His position is that you get a better result by capturing full frames (576 lines) and chucking away a field to end up with a 288 line non-interlace AVI ready for conversion.

My position is that its better to capture at 576 lines, use Smart deinterlacing to only deinterlace motion and then bilinear resize down to 288 lines. This is because you have more information in static areas (no combing in the source) to create the new 288 line AVI instead of just throwing every other line away.

I've found the result to be very good, but who is right?

Cheers,

Bolix

[mrtristan]

I don't think either of you is right or wrong, since it's entirely a matter of opinion.

Personally, I prefer your method.

I don't like to discard one of the feilds, because it can lead to noticeable choppy motion. But hey, maybe he's willing to sacrifice a little quality in order to delete a step.

[SatStorm]

It also depents on what you encode: mpeg 1 or mpeg 2?

The very few times I create VCD, I prefer capture both fields (576 Ver lines) and then let TMPGenc de-interlace, while it encodes.
When you grabb with Virtualdub, you have plenty of filters to use. Depending the quality of the source, you can choose what it does your job better.
Capturing from a Broadcast source like a TV program, gives you the best possible quality. For me is better to simply de-interlace even field. But capturing from VHS, is a totally different situation. Sometimes, on my old Tapes, Smart de-interlacing creates better results and sometimes no. Only a testing gonna show (thank God there are CD-RW on this planet!)

One reason I prefer mpeg 2 in generall, is exactly this issue....
You don't have to de-interlace . You just put your interlace source to the encoder and start encoding (interlace ouput, I am not a fan of progressive picture on mainstream TVs)

[bolix999]

So is there a point in capturing 576 lines if you're throwing 288 away? Does the capture card "prefer" that resolution?

I've done a bit more testing and it's hard to pick a winner. Even field only is a bit sharper and more noisy, smart deinterlace is smoother (though not so smooth it looks like a Bold and the Beautiful soft focus shot), less noise. The source is Video8 footage so I guess it's a bit hit and miss anyway.

I struggled quite a bit to get MPEG-2 interlacing to work (I had to swap the capture fields AND specify field A in Tsunami) and in the end I prefered the bang for buck on MPEG-1 lower res. MPEG-1 is a bit easier to muck around with as well.

I'm also comparing TMPEnc against CCE and so far TMPEnc looks a shade better at the bitrate I'm using (2000 CBR).

Cheers,

Bolix