Thresholding: A challenge for parallel processing

In the past the speed and low cost of parallel computation for image processing have often been emphasized. To give a sample of the possibilities of parallel processing, a typical task is assessed : image thresholding. We describe the performance of a sequential computer and of an existing parallel machine implementing algorithms which match the architecture of the machines. Performances are estimated in terms of the required processing times and it is shown that, although a great deal of arithmetic computation is involved, the greater the number of image elements, the faster the parallel machine becomes in comparison with the sequential machines.