How do I properly decomb a *resized* avi?

[Alcacia]

I have an avi with 2 problems:

1. it has the wrong aspect ratio - so it has clearly been resized.
2. there is a distinct 2-2 pattern of combed-progressive frames.

When I pass this avi through the Decomb filter, things do not get better, in fact they get worse...after going through decomb, the frames are still combed...and some of the combed frames have gone from the original "fine" combing to a "fat/thick" combing.

I am not sure, but guess that this could be because of the resizing? I tried to SeparateFields on the avi, and it didn't work. So if SeparateFields doesn't work, then I guess Decomb won't work, right?

I never thought about it before...but how exactly does SeparateFields work? I mean, does it always take every other row of *pixels* and assume that is a field? If that is the case, then SeparateFields will always fail on resized sources, and thus decomb will fail. It's funny that I've never seen any discussion of this issue.

Anyway, does anyone have any recommendations? Should I try to resize to either 480 or 576 lines *before* applying decomb? (I'll assume horizontal resolution is a non-issue). And which resize filter and parameters is best for cases like this?

Thanks for any feedback.

[Asrial]

If you have an AVI that has combs in it, they're hardcoded and cannot be removed easily.

You need the original source and it's recommended NOT to resize before deinterlace, but it doesn't break something to the point that there's a noticeable pattern the entire clip.

[Guest]

If you vertically resize an interlaced video without separating fields first, then you irretrievably ruin it for purposes of field matching and deinterlacing. Unless, of course, your resizing is by one-half. :)

[Alcacia]

Thanks guys, that makes sense I guess. I suppose also that any time you use any kind of frame-based compression (which is *always* the case with AVI's), then you also make decombing difficult or impossible, since the compression will result in a little blending of the fields...am I right?

Anyway, FieldDeinterlace ended up giving me a decent result, though not as nice as progressive reconstruction would have been.

[acrespo]

Originally posted by neuron2
If you vertically resize an interlaced video without separating fields first, then you irretrievably ruin it for purposes of field matching and deinterlacing. Unless, of course, your resizing is by one-half. :)

I usually capture at 720x480 then I deinterlace before resize to 640x480. I will have the same results if resize to 640x480 and deinterlace after?

[Guest]

Originally posted by acrespo
I usually capture at 720x480 then I deinterlace before resize to 640x480. I will have the same results if resize to 640x480 and deinterlace after? The results won't be identical pixel-for-pixel, but you probably wouldn't perceive any difference. As I said, it's only vertical resizing that needs to be avoided before deinterlacing.

[scharfis_brain]

@alcacia:

which is *always* the case with AVI's
but there are some codecs out there, that support laced compression:
huffyuv, xvid, mjpeg ....