NETRA - A HIERARCHICAL AND PARTITIONABLE ARCHITECTURE FOR COMPUTER VISION SYSTEMS

Computer vision is regarded as one of the most complex and computation ally intensive problems. In general, a Computer Vision System (CVS) at tempts to relate scene(s) in terms of model(s). A typical CVS employs algorithms from a very broad spectrum such as such as numerical, image processing, graph algorithms, symbolic processing, and artificial int elligence. This paper presents a multiprocessor architecture, called ' 'NETRA,'' for computer vision systems. NETRA is a highly flexible arch itecture. The topology of NETRA is recursively defined, and hence, is easily scalable from small to large systems. It is a hierarchical arch itecture with a tree-type control hierarchy. Its leaf nodes consists o f a cluster of processors connected with a programmable crossbar with selective broadcast capability to provide the desired flexibility. The processors in clusters can operate in SIMD-, MIMD- or Systolic-like m odes. Other features of the architecture include integration of limite d data-driven computation within a primarily control flow mechanism, b lock-level control and data flow, decentralization of memory managemen t functions, and hierarchical load balancing and scheduling capabiliti es. This paper also presents a qualitative evaluation and preliminary performance results of a cluster of NETRA.