A Topology Preserving Fuzzy Representation Technique For Pattern Recognition
N. Tavakoli · International Journal of Modelling and Simulation · 1990
A new representation approach is introduced that provides the advantages of both statistical and syntactical approaches by incorporating the topological information of an image in the features extracted from the image. The structural information is contained within the primitives rather than as relationships between them. The features represent a range of values that is a fuzzy rather than an exact representation, and thus eliminate the need for inexact matching. Applying this approach on a set of digitized characters bas yielded a 99.7% recognition accuracy.