Scalable Data Parallel Algorithms and Implementations for Vision.
Final technical rept. 1 Aug 94-31 Mar 95,
UNIVERSITY OF SOUTHERN CALIFORNIA LOS ANGELES INST FOR ROBOTICS AND INTELLIGE NT SYSTEMS
Pagination or Media Count:
Our research is about designing, analyzing and implementing scalable parallel solutions to problems in intermediate- and high-level vision. This is a difficult problem as computations are heterogeneous, symbolic and geometric in nature and use complex data structures such as lists and graphs. We propose a realistic model of distributed memory parallel machines which accurately models the features of a parallel machine. This includes the costs of communication-latency, impact of communication patterns on network congestion, available bandwidth and time for synchronization. We analyze the computation communication and control characteristics and the memory requirements of the vision algorithms. Our parallel algorithm achieve load balancing by dynamic redistribution of the tasks. We show the results of our approach in parallelizing the line finding problem on IBM SP-2 and a perceptual grouping step on TMC CM-5.
- Computer Programming and Software