LAUSR

Off-line text recognition is a branch of OCR. In Character recognition system, shape detection and feature mining is extremely vital part of the system. In most of the character recognition system of past researches they used zoning based features like directional features, distance based features, geometric feature etc., mostly used separately. In this paper, Binary Area Matrix calculation is introduced. The performance of the Binary zone area matrix is measured individually and with combinations of other existing features. India is a land of multi-script country with eighteen different scripts authorized by The Government of India Sridhar (Stud Linguist Sci 30(1):149–165, 2000). Many minor languages are available in Indian with its own scripts. Such one of the language is Saurashtra language which belongs to the Indo Aryan languages. We applied our method in this language for character recognition. For each letter of Saurashtra binary zone area matrix, zone entropy matrix, zone Euler matrix and Chain for weighted matrix features of 54 identical zones of the image in 0$$^\circ $$∘, 90$$^\circ $$∘, 180$$^\circ $$∘, 270$$^\circ $$∘ are extracted. The Performance of features is examined using SVM Classifiers. Combination having a binary zone area matrix, zone entropy and zone Euler can be classified into different text types. We obtained around 99% of recognition rate.