Montage: continuous presence teleconferencing utilizing compressed domain video bridging

This paper describes a multipoint video bridging architecture, called Montage, that is capable of composing and displaying an arbitrary number of full motion (30 frames/sec) compressed video streams. The system supports end-user-customization, where each end-point controls and customizes the video composition for his/her screen independently of any other participants. The coding and decoding schemes are extremely low delay, on the order of 1 msec coding and decoding, making it possible to support highly interactive applications. The Montage architecture consists of three components connected by a suitable broadband infrastructure, the encoder, the decoder and the video bridge. The Montage decoder has been implemented as a context switchable decoder. This is accomplished by performing all compositing and clipping manipulations on the incoming streams before decompressing them. Much of the functional flexibility of the decoder stems from the the Montage encoder which is unique in that it is capable of compressing full-size (640×480), entertainment quality NTSC video at 30 frames/sec and simultaneously producing three other reduced resolution compressed streams. As with the decoder, the encoder uses a single commercially available compression chip-set. The video bridge piece of the Montage architecture can take several forms, depending on the capabilities of the broadband fabric and the intelligence of the endpoints. Montage makes it possible to provide continuous presence video teleconferencing in a wide range of scenarios, including local area networks, PBXs, and wide area networks such as ATM