Computer recognition of arabic cursive scripts

Abstract

The main objective of this paper is to design an Arabic text recognition system. This system basically includes a segmentation stage in order to recognize typewritten Arabic cursive words. The segmentation process is completely based on tracing the outer contour of a given word and calculating the distance between the extreme points of intersection of the contour with a vertical line. At the output of the segmentation stage, the cursive word is presented as a sequence of isolated character contours. The recognition problem thus reduces to that of classifying each character. A set of Fourier descriptors are obtained from the coordinate sequences of the outer contour of each character. A topological classifier is also used to classify the stress mark over or under the character contour. A reject option is introduced by the classifier so that incorrectly segmented characters are detected. The developed system has shown a recognition rate of 99%. This result guarantees the efficiency of the used Fourier descriptors in discriminating between Arabic characters.