New Scientist has an interesting article about a technique to construct 3D information from two captured images - one taken with flash, the other taken without flash:
I was thinking - what if we applied this idea to video? It would be simple: attach a strobe light to a camera so that the strobe flashes for every other frame (frame rate of camera should probably be double that of normal to reduce movement between frames and to increase playback quality). We would then have a series of "real" 3D images.
Interesting things could be done: imagine watching a movie and being able to move the camera while it was playing to see the action from a different angle! This would be limited to shallow angles, since the depth is only per pixel (x/y, not z), but it would still be very cool.