A Novel approach for Segmentation of Typewritten Gurmukhi Script

Abstract

Demand of OCR (Optical Character Recognition) based frameworks have expanded definitely because of their noteworthy focal points into continuous applications. Segmentation assumes a significant job in optical character recognition of typewritten Gurmukhi script. Segmentation of typewritten documents is a challenging task due to the presence of skewness, overlapping, and degraded characters. Improper segmentation can hamper accuracy of character recognition. In this paper, we are proposed a new technique for line segmentation by modifying A-star algorithm and combining it with strip-based projection. Character segmentation technique is also proposed on the basis of horizontal and vertical projections combining with the aspect ratio of characters. We have accomplished accuracy of 94.28% and 99.78% with pixel count for line segmentation and 95.70% for character segmentation.