Computer vision and videoconferencing: flirt or long-lasting love?

Photograph of Emanuele Trucco

Emanuele Trucco


The last 10 years have seen a number of efforts towards the development of videoconferencing systems based on computer vision and image processing.

Recent examples include the VIRTUE system, the Coliseum system of HP Research Palo Alto, and the US Tele Immersion Initiative. The algorithmic power of contemporary computer vision is certainly seductive for advanced, image-based communications; but can computer vision deliver in such a demanding, market-driven domain? We approach this question from the point of view of applied computer vision research. On the way to an answer we discuss some recent techniques and results, from our and other laboratories, especially in the area of image-based rendering and specifically view synthesis. Key issues here are the generation of very accurate disparity maps for the human body in a videoconferencing set-up, and the data-driven vs model-based dilemma.
Tuesday 12th October 2004, 14:00
Robert Recorde Room
Department of Computer Science