Automatic Regonition for Old Arabic Manuscripts

Segmentation-free approach with texture analysis forms the basis of object recognition and classification in several domains, one of these domains is old Arabic manuscript recognition. The basis of this approach is to analyze the texture structure of the document image independent from its content, author writing style¿etc. The main objective is to propose an automatic system for recognizing and annotating the undated old Arabic document manuscripts in 4 different ages: Contemporary (Modern), Ottoman, Mamluk, and ages before Mamluk. The system has been tested by 129 different Arabic manuscripts documents taken from Dar Al-Kotob Al-Masria Library and Leiden University Arabic manuscript collections. The results yield to 95.7% correct annotation for dated Arabic manuscript and 92.3 % for undated Arabic manuscripts.