Accelerated recognition of handwritten Urdu digits using Shape Context based gradual pruning

Prototype based object recognition requires measuring similarities between the test object and the prototype categories. When multiple instances per object are stored in the prototype set, this task becomes computationally expensive. To increase the efficiency, pruning techniques are used. In this work we have presented a gradual pruning approach based on the dissimilarities between the test object and the objects in the prototype set. A mathematical expression has been derived analytically for the savings in computational time. This approach is applied to the task of hand written Urdu digit recognition using Shape Context. Compared to the classical (non-pruned) and step pruning approaches, our gradual pruning based method is found to be faster without compromise on accuracy.