Parallel algorithm in image morphology

Abstract

Summary form only given. A parallel specification has been obtained for the opening and closing operations in image morphology in one and several dimensions. This has been done for the usual image case as well as in the wraparound case. The role of mathematical morphology in determining other multidimensional algorithms has also been identified. For instance, Mellin convolution is easily found by first utilizing the dilation operation. All the algorithms have been expressed using data-flow type diagrams and allow straightforward use of fault-tolerant methods. The unified signal algebra approach to parallel algorithms and architecture design has been used in developing all the procedures reported. It permits the design of data-flow systolic and wavefront array architectures.<>