Heuristic approach to the recognition of printed Arabic script

Abstract

A new segmentation-free method, called N-markers, is proposed for machine recognition of the Arabic printed texts. The contribution aims at the optical character recognition of printed texts, like books and journals of good quality, usually typeset in so-called Naskhi font. The focus of attention is shifted from the recognition of multifont texts to that of single Naskhi font, taking, however, into account shape variations originated in different typesetting workshops, and the intensive presence of the ligatures in normal printed texts. The proposed method is a mixture of global and structural approaches and is related to some early ideas of the optical character recognition (OCR) of the isolated Roman characters.