Title :
Thinning in a distributed environment
Author_Institution :
Dept. of Comput. Sci., Calgary Univ., Alta., Canada
Abstract :
In the digitization of survey maps where the components are sparse but occupy large areas, a serial thinning algorithm implemented on a distributed environment can yield a better speedup than is possible with other forms of parallelism. A distributed algorithm based on contour generation is described. A component is divided into rectangular sections and assigned to different processors. A block-resume synchronization mechanism is examined. The different contour configurations at the border of a section are identified. The amount of communication between neighboring sections can be kept to a minimum by chain code representations. The proposed synchronization mechanism has been incorporated in the contour generation thinning algorithm and has been simulated on a Sun/4 workstation. For images such as contour maps, the extra overhead needed for synchronization is not significant
Keywords :
computerised pattern recognition; computerised picture processing; distributed processing; Sun/4 workstation; block-resume synchronization mechanism; computerised pattern recognition; computerised picture processing; contour generation; contour maps; digitization; distributed environment; distributed processing; serial thinning algorithm; survey maps; Computer architecture; Computer science; Costs; Distributed algorithms; Distributed processing; Iterative algorithms; Iterative methods; Parallel processing; Partitioning algorithms; Skeleton;
Conference_Titel :
Pattern Recognition, 1990. Proceedings., 10th International Conference on
Conference_Location :
Atlantic City, NJ
Print_ISBN :
0-8186-2062-5
DOI :
10.1109/ICPR.1990.118195