Real-time distributed and parallel processing for MPEG-4

MPEG-4 is currently being developed by MPEG to specify the technologies for supporting current and emerging multimedia applications. Because of its object-based features and flexible toolbox approach, it is much more complex than previous video coding standards. We believe that software-based implementation on parallel and distributed computing systems is a natural and viable option. In this paper, we describe such an approach on the MPEG-4 video encoder using a cluster of workstations. We propose to use hierarchical Petri nets as a modeling tool to describe the temporal relations and time constraints among various video objects at different levels. This would allow us to perform scheduling with a guarantee of synchronization among multiple objects. A dynamic shape-adaptive data parallel approach is used in the spatial domain for further speed-up gain. Our preliminary results indicate that real-time MPEG-4 encoding using distributed and parallel computing is achievable.