A Hybrid Method for Three Segmentation Level of Handwritten Arabic Script

Fadoua Samoud, Samia Maddouri,  and Noureddine Ellouze
Image Processing and Pattern Recognition Lab, National School of Engineers of Tunis, Tunis

 

Abstract: The main theme of this paper is the segmentation of handwritten Arabic script into blocks, connected components and characters using a combination between Hough Transform (HT) and Mathematical Morphology (MM) tools. We start by a segmentation methodology of a complex document into its distinct entities namely handwritten components. Each extracted handwritten blocks are then segmented into sub-words as a main specificity of Arabic script. Finally a character segmentation method is presented. For each segmentation step, some concepts are needed such as dynamic kernel and Harris corner detectors. The proposed method is tested on the CENPARMI Arabic check database and on the IFN/ENIT database. We present a concept for automatic evaluation of the results, based on label tools for the different parts of used documents. 

  

Full Text