Radon transform and dynamic programming for the Persian handwritten zip code recognition

Abstract

Pattern recognition is one of the major research areas in computer sciences. Optical character recognition OCR as one of the pattern recognition topics has specifically attracted the interests of many researchers. This paper presents a method for offline Persian numeral character recognition. It uses Radon transform for feature extraction and a combination of dynamic programming with K-nearest neighbour KNN for its recognition task. For the simulations, the Persian postal zip code database of 800 samples has been employed, providing a recognition accuracy of 94.1%.