The topology of digital images (Q1205567)

In digital imaging one wishes to represent a ''subject'', a set \(P\) in a topological (usually Euclidean) space \(S\), by its ''digital image'', a set \(A\) in a discrete space \(E\). In the applications \(E\) models a set of on/off picture elements. In order to recover topological information about a subject from its digital image, the set \(E\) must be equipped with a ''decoding structure'', relating it to the topology of \(S\). In the original version of this problem \(S\) is a rectangle in the real plane and \(E\) consists of those points in \(S\) with integer coordinates, but higher dimensional applications are increasingly important in medical imaging (now up to \(n = 5\)), and there are other possibilities even for \(n = 2\) (for example, a hexagonal rather than rectangular arrangement of picture elements). The author develops a mathematically elegant general theory which provides a topological setting that includes all of the cases mentioned in the previous paragraph. His decoding structure is constructed as follows: A ''fenestration'' of \(S\) is a collection of disjoint, non-void, open subsets, called ''windows'', whose union is dense in \(S\), with each window corresponding to a point of \(E\). Every fenestration of \(S\) determines a family of quotient spaces of \(S\), called ''grids'', in each of which \(E\) is embedded as a dense discrete open subspace. Each grid can serve as a decoding structure, but there is a unique minimal grid. It satisfies \(T_ 0\) separation, but not \(T_ 1\) in general, and is constructed as the smallest element in a natural partial ordering of grids. There are extensive results on connectedness properties. In real computer applications, the set of picture elements must of course be finite. Although the results in this paper shed much light on finite topological spaces and their connectedness properties, the point of view here is broader: much of the theory of grids is developed in the more general context of spaces having a dense set of isolated points. As one example of a result of more general topological interest, any semiregular minimal grid can be construed as a collection of filters on \(E\) with an inherent Wallman-type topology. This paper is part of a collection of related papers; see Guest editors' preface to special issue on digital topology by \textit{T. Y. Kong}, \textit{R. Kopperman} and the reviewer [Topology and its Applications 46, 173-179 (1992, Zbl 0762.54035)].