On Fri, 24 Aug 2007 21:58:55 +0200, Derek McTavish Mounce
<[EMAIL PROTECTED]> wrote:
Forgot this. More of a question on what you mean exactly:
An NLE that deals
with TV material (not just cinematic stuff) ought to be able to
preserve interlacing throughout the workflow, no matter how many
effects or transformations you throw in.
It is absolutely not possible to apply a transformation other than
simple
translation to interlaced video and maintain the interlacing. If you
scale or rotate the image, you're scaling/rotating the fields which
completely defeats the purpose of interlacing. You must deinterlace.
There's not a program out there that can maintain proper interlacing once
you scale, rotate, or apply fx.
Perhaps though I'm misunderstanding, in which case please explain
further. :)
Maybe, maybe not.
You do not have to deinterlace in the common meaning: Merging fields
into frames, thus reducing 50i to 25p. To maintain the temporal
resolution, one has to do the opposite: separate the fields, and
process them separately. They are separate images, after all.
Here is how I think it should go:
1) Split out the fields, putting them after one another on the timeline.
2 a) Line-double or interpolate them up to full vertical resolution
or
b) Make the rendering pipeline treat the fields as very wide
non-square pixels.
3) Transform, rotate, translate, filter, whatever the fields...
4) Merge back the fields to frames, if/when appropriate.
Since field 1 and 2 are separated on the timeline, all the pixels
in the final field 1 will come from the source field 1, no matter
the deformations that have been applied.
The pixels may need stretching or interpolation to fit into their
new positions, but keeping the fields apart is just a matter of
combing them out and putting them in their correct temporal position
on the timeline.
A capable NLE should do that.
--
Herman Robak
_______________________________________________
Cinelerra mailing list
Cinelerra@skolelinux.no
https://init.linpro.no/mailman/skolelinux.no/listinfo/cinelerra
