Parallel reduction of digital sets

From MaRDI portal





In their book on perceptrons [Perceptrons. An introduction to computational geometry (1969; Zbl 0197.437)], \textit{M. Minsky} and \textit{S. Papert} concluded that it is not possible for a parallel computer to decide any nontrivial topological predicate. Since the topological properties of digital sets in binary pictures are sometimes relevant (e.g., for the automatic testing of printed circuits), the barriers imposed by this negative conclusion are usually circumvented in picture processing by thinning the content of an image, i.e., representing the black regions by topologically equivalent ``thin graph-like structures. This is achieved by successively removing black points from the set while retaining its topological properties. The analysis of graph-like structures by sequential processes is much easier than the topological analysis of the original set. In the present report a new concept of perfect points is introduced and its application to parallel thinning is investigated. It is possible to characterize the irreducible sets obtained by the application of the new method. Thinning algorithms can of course lead to irreducible sets having large clusters of interior points. Such structures are certainly not ``thin in any reasonable sense. It is shown that such unwanted structures will not occur if a very simple and natural condition is assumed to hold.











This page was built for publication: Parallel reduction of digital sets

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3787527)