On separation of english numerals from multilingual document images

Abstract

For Optical Character Recognition (OCR) of bilingual or multilingual document containing text words in regional language and numerals in English, it is necessary to identify different script forms before running an individual OCR of the scripts. In this paper, an attempt is made for separation of English numerals at word level from bilingual and trilingual documents representing Kannada, Devnagari, Tamil, Odiya and Malayalam scripts by using discriminating features such as aspect ratio, strokes densities, eccentricity, etc. as a tool. The k-nearest neighbour algorithm is used to classify the new word images and the algorithm is tested on 6000 sample words with a five fold cross validation test. The algorithm is robust with respect to font styles, sizes and noise. The results obtained are quite encouraging. © 2007 ACADEMY PUBLISHER.