Pre-processing for Improving the Recognition of Mobile based Gurmukhi Text Recognition System

Abstract

With the advent of technology and portable computing devices such as mobile phone new trends of research have emerged. The approach to character recognition is no longer limited to desktop scanners. Text detection and recognition in document images using mobile device is becoming popular due to the increasing interest in on-demand processing of documents. There are many ongoing researches in this field. In this paper image pre-processing techniques are discussed to improve the accuracy of the mobile based Optical Character Recognition (OCR) system. Pre-processing is integral part of any OCR System as it is applied to image before passing it to the OCR. Process involves image capturing and cropping, rotation, gray conversion, binarization, shadow removal, skew correction, removal of noise, and light variations of the image. To verify the accuracy, OCR application is tested with 50 document images before and after pre-processing. Experimental results show that pre-processing techniques significantly improve the OCR accuracy while decreasing the overall computation time.