Image Enhancement Processes for Myanmar Printed Text Recognition

Abstract

Image enhancement processing step is the basic crucial part of the OCR system as the recognition accuracy of OCR systems greatly depends on the quality of the input text image. Most of the OCR systems, especially Myanmar OCR systems are little effort for preprocessing or image enhancement process. And Myanmar glyphs are complex in shape; combine with many circular, dots and line and pixel distribution to form this scripts are not uniform. Therefore, making the input image to be better is important for our native OCR system. In this paper we propose image enhancement processes suit that, through reducing noise, separating text and background, skew detection and rotation, improves the accuracy of the OCR System for Myanmar printed text. These steps can produce a refined image that is ready for the segmentation or character extraction process of the OCR system.