Abstract
If the pre-processing phase, in optical character recognition systems, is the heart of the recognition process, the segmentation stage is the "aorta" of this heart. This paper introduce a reliable segmentation technique for Arabic handwritten script. Number of techniques like, script height, character width, pen thickness and word/subword gaps are used to design an efficient segmentation algorithm. The algorithm performs diacritics removal, word/subword segmentation, ascender characters segmentation, descender characters segmentation and finally embedded characters segmentation.