HIGH ACCURACY RECOGNITION ALGORITHM OF ARABIC CHARACTERS

Abstract

This paper presents a fast high accuracy recognition algorithm for typewritten Arabic characters. The procedure of recognition consists of two parallel stages. The first one is devoted to recognize the main body of the charcacter where the outer contour of the character main body is obtained and a limited number of Fourier descriptors are derived from the resulting contour. The second stage utilizes the topological features to classify the stress mark(s) if any exist, the extraction of topological features depends on a new technique called Pixels' Numbers and Position (PNP) of stress marks. Combining these two interrelated stages gives high recognition results while maintaining computational simplicity.