Recognition of handwritten Urdu digits using Shape Context

Many approaches and techniques have been explored with respect to the problem of digit recognition. In this paper we present a fresh approach towards matching and recognition of hand written Urdu digits using the novel descriptor for shape matching, the 'Shape Context' recently proposed in University of California, Berkeley [1],[2]. Our representation for a hand written digit is a discrete set of n points sampled along its border. For each of these points, the shape context is a histogram of relative positions of the n-1 remaining points. We have taken the criterion of similarity between two instances as the weighted sum of cost of matching shape contexts and bending energy, which is the amount of work it takes to transform one instance to another. We found the technique to be effective with zero percent error on the 28 test digits.