Reconstructing the 3D geometry of scenes captured through a camera has been a subject of computer vision research for many years. The theoretical mathematics of calculating the 3D shape of static scenes using images taken from different viewpoints is known, and many practical implementations of this are used today. However, reconstructing scenes containing dynamic content remains a far greater challenge.
We address this challenge and work on obtaining the 3D geometry and motion of dynamic objects captured by a single digital video camera. We not only recover the exact geometry and motion of the captured object, but work on driving computer graphics systems to create models which can be interacted with, rendered in different styles, or vary from the original to create a new individual model of the same type.
The projects we have been working on, past and present, are linked to on the right.
