Multiple narrow-baseline system for immersive teleconferencing

Abstract

An important aim of immersive teleconferencing systems is to create realistic 3D virtual views of remote conferees. Hence, systems should be able to deal with hand gestures as well as occluded areas in reference images required in derived views. The quality of such derived views is dependent not only on the analysis and synthesis process but also the multiview camera set-up. Often the popular convergent wide-baseline stereo approach aspires to achieve too much through a single camera pair: maximum information and reliable disparity maps. We identify how this dichotomy leads to problems in the analysis and synthesis process, often leading to a restrictive system specific solution. We then define a new approach, a multiple narrow-baseline set-up, designed to overcome the limitations of the wide-baseline set-up, being modular, both in terms of system requirements as well as algorithmically, and scalable, with respect to the number of conferees.