Multi-Stream Management for Supporting Multi-Party 3D Tele-Immersive Environments

Three-dimensional tele-immersive (3DTI) environments have great potential to promote collaborative work among geographically distributed participants. However, extensive application of 3DTI environments is still hindered by problems pertaining to scalability, manageability and reliance of special-purpose components. Thus, one critical question is how to organize the acquisition, transmission and display of large volume real-time 3D visual data over commercially available computing and networking infrastructures so that .everybody. would be able to install and enjoy 3DTI environments for high quality tele-collaboration. In the thesis, we explore the design space from the angle of multi-stream Quality-of-Service (QoS) management to support multi-party 3DTI communication. In 3DTI environments, multiple correlated 3D video streams are deployed to provide a comprehensive representation of the physical scene. Traditional QoS approach in 2D and single-stream scenario has become inadequate. On the other hand, the existence of multiple streams provides unique opportunity for QoS provisioning. We propose an innovative cross-layer hierarchical and distributed multi-stream management middleware framework for QoS provisioning to fully enable multi-party 3DTI communication over general delivery infrastructure. The major contributions are as follows. First, we introduce the view model for representing the user interest in the application layer. The design revolves around the concept of view-aware multi-stream coordination, which leverages the central role of view semantics in 3D video systems. Second, in the stream differentiation layer we present the design of view to stream mapping, where a subset of relevant streams are selected based on the relative importance of each stream to the current view. Conventional streaming controllers focus on a fixed set of streams specified by the application. Different from all the others, in our management framework the application layer only specifies the view information while the underlying controller dynamically determines the set of streams to be managed. Third, in the stream coordination layer we present two designs applicable in different situations. In the case of end-to-end 3DTI communication, a learning-based controller is embedded which provides bandwidth allocation for relevant streams. In the case of multi-party 3DTI communication, we propose a novel ViewCast protocol to coordinate the multi-stream content dissemination upon an end-system overlay network