New rules to enhance the performances of histogram projection for segmenting small-sized Arabic words

Abstract

Off-line Arabic segmentation has been a popular field of research. It still remains an open problem for discussion. In fact, the challenging nature of Arabic writing which increases the complexity of recognition and segmentation task has attracted the attention of many researchers. This paper proposes and investigates an enhanced algorithm based on the vertical histogram projection and some rules to segment Arabic words with small size. These rules are based on not only the structural characteristics of Arabic language, but also on the baselines positions and their relation with the characters. Our approach aims at cooperating together the segmentation method based on histogram projection and the contextual topographies of Arabic writing in order to improve the segmentation rate. Thus, we use the vertical histogram to detect the preliminary segmentation points and some other rules to find real segmentation points. The proposed approach has been tested with Arabic Printed Text Image Database (APTI). Actually, promising results have been obtained. Compared with the previously-proposed approach, our algorithm gives better result if applied on smaller size.