NETWORKING FOR IMMERSIVE TELEPRESENCE: ARCHITECTURES AND PROTOCOLS- A CASE STUDY

Immersive telepresence allows an observer to view a remote scene from any viewpoint of choice and thus give a sense of presence, as opposed to conventional video viewing where the user can only view content from the viewpoint of a single camera. We specifically consider a depth-based rendering approach for enabling telepresence. In this paper, we explore various networking architectures in which telepresence can be enabled over the Internet using this approach. From these architectures, we derive a set of requirements for describing and conducting a telepresence session. Based on these requirements, we present a session protocol that draws upon the features of RTSP and SDP and extends them. Various media streams from a single camera viewpoint are aggregated as a view group and the concept of this view group is used to support the various architectures. Both unicast and multicast configurations, with central or distributed servers, are supported by this protocol. Finally, we present the overall end-to-end architecture used in our current implementation. 1