A Novel Technique of Optical Character Recognition of Printed Text for Multi Languages

Multi-scripts recognition systems are requisite in the countries like India where multi-languages are spoken in numerous states of the country. Multi-Scripts Recognition is a demanding problem and research work for expansion of optical character recognition scheme for bi-scripts and multi-scripts is in infancy. Here in presented work a multi-script recognition system is proposed for the English and Punjabi scripts. For recognition the image is processed through basic steps of OCR like pre-processing, segmentation, feature extraction, correlation calculation and classification. After binarization and noise removal of the test image, it is segmented using line segmentation, words segmentation, and character segmentation technique of proposed algorithm. The lines of both the languages are segmented using the horizontal projection profile of the image. The words segmentation of the Punjabi language is very easy as compared to the English language due to the fact that all characters of a Punjabi word is connected through a connecting line and two simultaneous words are separated by blank space. After the segmentation, the number of holes in image is calculated and then the correlation of this character is calculated with the particular group of trained database. And then the decision is made on the basis of the highly correlated image in the database. Grouping of the database is done to reduce the correlation calculation time for whole database. In last system efficiency is calculated by using the test images of various sizes. Experimental results show the high accuracy of the proposed system.