Segmentation versus segmentation-free for recognizing Arabic text
M. G. E. Allam · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 1995
Arabic character recognition faces many technical difficulties, but the most challenging problem is the cursive characteristic of Arabic text. The present work explores two different approaches to solve the cursive problem. The first approach depends upon preprocessing to segment connected characters. The other approach is a segmentation-free technique where whole portions of connected characters are recognized without prior segmentation. Both methods employ hidden Markov models for classification. The present paper investigates the robustness of the two techniques and suggests how to combine them so that the weakness of one is compensated for by the strength of the other.