Real-time Video Processing for Tele-presence

Tele-presence systems aim at creating the illusion that all participants at a remote location are in the same virtual room. As a new emerging technique, it is faced with many challenging problems, one of which is automatic, real-time foreground/background video segmentation, or called bi-layer video segmentation. The current method developed in our laboratory  for the bi-layer video segmentation problem is based on the fusion of infrared and color video. The method improves on previous algorithms by making the system independent of changes in ambient lighting and dynamic background, and it achieves the goal of automatic video segmentation in real time.

We are now looking for a new Masters student that will be fully paid for two years to continue the development of this system using this time a real-time time of flight camera. Knowledge of computer vision and CUDA is mandatory.