Neighbourhood-based vision systems

Abstract

The problem presented in this paper is how to find similarities between digital images useful in design cybernetic vision systems. The solution to this problem stems from a neighbourhood based vision system. A neighbourhood is viewed in the context of a covering of a visual space defined by tolerance relations. A consideration of neighbourhoods and tolerance classes leads to a highly practical tolerance near set approach in vision systems. The contribution of this article is an algorithm for finding tolerance classes, a new measure for quantifying the similarity between tolerance classes, and a practical application of the tolerance space approach.