I'd love to know about this too for an upcoming project - I may be way off beam, but I've heard the concepts of "Difference" and Difference squared" used in this sort of context.

I'll do some reading when I have time, but maybe someone who knows more can add something ....

Yeah Peter,
I saw a refence to the difference squared in the post on depth of field.
I think it was Sony Dennis that had a post, but I could not get to the link he posted. Hopefully someone will give more info.

Just an idea here, possibly as intensive as bezier masks but hopefully a bit faster. Export the clip as a frame sequence, erase the background in each image individually and reimport the images as a sequence. I am only recommending this way because it seems as time consuming as the bezier masking will be.
I think the difference squared that some of you guys mentioned is something that a camera records. This might be sci-fi, but I heard that there is technology where in addition to the usual the information about the distance between objects in a frame is also recorded.

Peter is referring to Vegas' Difference Squared compositing mode.

Rob

Wouldn't it be easier to just re-shoot the dancing? Of all things, dancing would be hard to mask (versus a person sitting relatively still in front of a camera).

You may also try "Ultra" from Seirous Magic. They use a different technology and seem to handle diverse backgrounds (but not sure if they can handle "that" diverse).

i'd just use bezeir masking...it may take a while but it's probably the simplest way to do something like that, i did it with a skateboarder going down a rail and it turned out way good. Well later

-Will

I think this "demo" might give you some ideas:

Cat Demo

Hey Edge...
if you DO use this (fabulous) difference technique would you let us see the results?
The cat demo is awesome (and very inspiring).
v