Personal Identification Based on Texture Analysis of Arabic Handwriting Text

This paper presents a personal identification method based on texture analysis of Arabic handwriting text. The method includes three steps. In the first one, the text image is processed in order to obtain a normalized and textured text. This is done by detecting and correcting the slope angle, normalizing space between lines and words, and filling margins. In the second step, the text image is analyzed by a set of 16 Gabor filters. A signature vector composed of 32 features is obtained. In the third step, signature vectors are compared to each other in order to estimate text similarity. The method has been tested on two databases: the first one contains printed texts using 20 different characters fonts and the second one contains samples of handwriting text written by 22 persons. The obtained results show that the method is able to recognize either printed text or handwriting text