Development Platform for Parallel Image Processing

This paper describes a development distributed platform with client-server architecture that allows developing parallel primary image processing on a cluster with variable number of workstations. The principles of the software and hardware architecture of this platform are presented underlining the versatility and the capacity of adaptation to a specific application. Experimental results show that for a realistic image processing application performances are accurate and consequently the core of the architecture forms a powerful basis for automatic parallelization of a wide range of image processing software. Key-Words: distributed platform, scheduling, parallel processing, image segmentation and rebuilding