Have you tried using OpenCV for this? You can do a background subtraction on a video of the person in the actual location to get all the foreground pixels corresponding to her/him. And then replace the corresponding foreground pixels of another video (traffic, restaurant, etc.) with these pixels. It would be in real-time, but maybe not so clean.
Hope this helps.
On Mon, Mar 26, 2012 at 10:05 PM, Ken W. Grant
<ken.w.grant@xxxxxxxxx> wrote:
I'm looking for a way to film people reading words and sentences and then place them in
different background scenes, like a busy street with pedestrians and cars going by or a
cafeteria with a lot of activity. Does anyone know a good software program where these two or
more source films can be merged and flattened so that the filmed talkers appear to be part of
the background scene? It doesn't have to be perfect like Lucas or Spielberg, but I'd like it to
look somewhat realistic.
Thanks in advance,
Ken
kenneth.grant@xxxxxxxxxxxx
www.avspeechlab,com