Contextual classifiers are being developed to exploit spatial/spectral content of a pixel to achieve higher classification accuracy. Contextual classification requires large amounts of computation, so special hardware is of value. One parallel processing system designed for image processing is the CDC Flexible Processor Array, the basis for the discussion in this paper. A simulator for the CDC Flexible Processor Array has been developed for program testing, debugging and timing. The simulated timings are presented in this paper. For comparison, the same algorithms have been run on a PDP-11/70. These timings are analyzed and discussed for context neighborhoods of size three and nine.
Date of this Version