PDA

View Full Version : handling hybrid film/video HDTV caps


phædrus
12th March 2007, 16:53
To avoid endless hassles and compromises that usually weren't wholly satisfactory, the old rule used to be "keep it interlaced."

Now with the problem of resizing to DVD resolution, all those hybrid problems resurface again. :mad:

How does software like HDTV2DVD handle this problem? Does anyone know? I'm curious.

I had this crazy idea of somehow taking 1080i and throwing away every third and fourth line, but keeping the remaining ones interlaced. then you'd have 540i. Trim 30 lines off the top and bottom and you'd have 480i. Fiddle with the borders on the sides and you could maintain correct aspect ratio for the part of the picture remaining after the trimming. It probably wouldn't look the best, but at least it wouldn't be jerky. Like I said, crazy talk. I wouldn't know how to implement it in script or code anyway.

I just capped a documentary on PBS that looks to be hybrid. I'm not going to make a big deal about it -- I'll just throw away every other line, keep the video framerate, and let the filmed portion play with some repeated frames (fields). But if I DID want to make a big deal out of it, are there any new strategies or tools out there that have developed since I gave up trying to force hybrid into progressive frames a long time ago?

I searched "hybrid" in this forum and did not find much. Apparently this isn't a burning issue.

Mug Funky
13th March 2007, 06:38
you could search for "ntsctools", but bear in mind it's not complete and still a little buggy. there's a better version on my hard disk, but i haven't touched it in a while.

more specifically, there's a way with these functions to conform all possible NTSC timebases (24, 30, 60) to one of your choosing, using motion-compensation to keep things smooth. so you could go 24p without the 30p bits going jerky.

however, it'll be mind-crushingly slow on HD footage. best just use a motion-adaptive deinterlace and resize down - progressive stuff wont get hurt, 3:2 patterns will be maintained (at least to a good enough level for SDTV) and you can encode interlaced.

so leakkernelbob it, or if that's not precise enough, tdeint (with matching for the 3:2 parts) should work fine.

phædrus
18th March 2007, 17:57
MF, thanks for the tip. I took a look at the documentary in question, and it looks like, apart from the pure video framerate stuff, some kind of telecine of material that had previously been badly IVTC'd. So it is seriously messed up and I'm not going to try to straighten it out. I'll just take the even fields, resize, and go with that.

However, not all hybrid material is going to be this hopelessly bad. I'm thinking of a promo I saw for a Cousteau special coming up on PBS. Likely some of that will be film and some video material. And probably the telecining on the filmed portions will be expertly done. So, as per your advice:

"so leakkernelbob it, or if that's not precise enough, tdeint (with matching for the 3:2 parts) should work fine."

Can you explain to me, in practical terms, what this will mean for the output? What does leakkernelbob or tdient do that FieldDienterlace(blend=false) does not do, in layman's language?

I got into this a couple of years ago when I had some hard telecined anime that I wanted to deinterlace and RePAL, but that was a single project and I had to have help from the kind forum members here. I didn't really know what I was doing -- I did try leakkernelbob and tdeint, and I can't remember what I finally used. Other than that, I've tried to stay away from IVTC and hybrid in particular.