Recognition of Arabic printed characters using structural features and contour methods

ABSTRACT

In this paper, a recognition system of Arabic printed text is presented, using structural features and contour methods. The structural features method is based on the detection of holes and concavities in four directions [up, down, left, and right] and the secondary characteristics of this method are used to differentiate between the characters of each class and to reduce the confusion encountered in the system. The contour method divides each character into a main part and numbers of secondary parts and, by the result of this division, four classes can be received of which each class can be deal with many characters, and to recognize these character, other features such as Cross-Points [Loop], Secondary Parts, Similarity of Secondary Parts, Bottom-Up Density Ratios[BUDR], Left-Right Density Ratios[LRDR], and Horizontal–Vertical Orientation [HVO] have been used. In this paper, the two methods are tested on isolated characters and on a block of text which has 266 characters. The tests are also passed on multi font and multi size. The experimental results showed a relatively high accuracy of recognizing Arabic characters by two methods. The recognition rate of structural features method was 90.01% and by contour method, it was 98.02%, giving a better result of the structural features method.