Visual attention based information culling for Distributed Virtual Environments

摘要

The central goal in Distributed Virtual Environments (DVEs) is to provide an experience of the shared world that is perceptually plausible for the immersed user. This has to be achieved in the face of network issues such as bandwidth limitations and latencies which make total synchronism impossible. Techniques have been devised to help; some optimise network use by limiting what is transmitted, others try to disguise the effects of network 'glitches' at the user's client machine - for example smoothing policies. To date such techniques tend to be system-based rather than user-oriented. Yet there is an active body of psychological research, for example in visual attention, which has successfully been employed by the graphics community to yield better perceived results for a given resource. Immersive DVEs are even more critically dependent upon the users' perceived experience. We may expect therefore that such 'psychologically' oriented approaches have even more to offer here. In this paper a visualattention model which exploits the peculiarities of the human visual system is presented. It is based on previous work, and on a series of carefully designed experiments which are used to guide the implementation of the model and to design architectures for DVEs. The proposed architectures are then tested using different DVE experimental applications, which include a highly populated virtual city. The results demonstrate that the characteristics of the human visual system can be exploited to improve network usage and generate a more perceptually plausible environment in the face of incomplete synchronisation.