Evaluation of Skeletonization Methods for Arabic/Farsi Handwriting Recognition

A goal-directed evaluation of four skeletonization methods for recognizing Arabic/Farsi handwritten words is presented. Two of which are well established and the third one has recently been reported. The fourth one is a novel parallel method proposed by the authors. This new method with the addition of a postprocessing step is shown to be the best taking into the consideration the fact that the handwriting recognition system using it yields higher recognition rates. The recognition is based on continuous hidden Markov models (CHMMs) with structural features.

Keywords:

Skeletonization , Thinning , Postprocessing , Structural Features , Arabic/Farsi Script , Handwriting Recognition , Evaluation