Vector morphology and iconic neural networks
Stephen Wilson · IEEE Transactions on Systems Man and Cybernetics · 1989
Mathematical morphology involves the geometrical analysis of shapes and textures in images. Methods of generalizing morphology are presented, and it is shown that all common image-based operators are instances of two fundamental operators where voting logic is at a pivotal point. Another generalization leads to vector operators. A sequence of vector morphology operations is similar to a multiple-layer iconic neural network. In morphology, a new operator called a weighted rank order filter becomes apparent. It is noted that massively parallel, bit serial computer architectures are the most effective way to realize the various operations discussed.>