Character Recognition Using Mathematical Morphology

Abstract

The automatic recognition of characters dates back to several decades, the reason being the large diversity of applications which make valuable use of recognition techniques. This paper proposes the recognition of a set of handwritten digits using mathematical morphology. Certain features that can be used to clearly distinguish one digit from the other are extracted. A decision tree has been developed to aid in the classification process. At each node a morphological operation is applied on the incoming digit, and the results dictate what the next step in the sequence should be. The method has been applied on a large sample collected from a diverse set of individuals. The recognition rates achieved are appreciable and substantially support the techniques discussed.