Three Dimensional Video : The Video of the Future

So far videos have been inherently two dimensional. During the process of recording a movie, the third dimension - the depth - is lost. When we watch a video, we are able to perceive depth because many depth cues are still present in a video like motion parallax, relative size, occlusions, etc. These depth cues are unfortunately ill-defined for robust processing by a machine, thus if we are interested in automatic high quality video manipulation like injection, replacement and deletion of objects in a video, or the generation of novel viewpoints, we have to know the depth map of the scene. Having recovered the three-dimensional structure of the scene in sight, we have turned a flat two-dimensional video into a three-dimensional experience, where the viewer is an immersed observer of the scene.

A simple sketch of our lab

Our research in this matter is described on the following web page:

Visual Space and Computational Video Geometry

Some demonstration videos are available at: Talking Heads Sequence

For further information please write to Jan Neumann