An Improved Enhancement Of Degraded Binary Text Document Images Using Morphological And Single Scale Retinex Operations

Abstract

This paper proposes a novel enhancement algorithm for degraded binary document images captured using a low end cell phone camera. The algorithm starts with colour to grey scale conversion followed by contrast stretching. Then single scale retinex enhancement is performed to increase the contrast, reduce global variance and improve the background uniformity of the image. Otsu's binarisation method is then used to dichotomize the image and finally, morphological dilation is performed to preserve stroke connectivity of the document characters by bridging any gaps resulting from the thresholding process. Computer simulation experiments have been used to demonstrate the effectiveness of the proposed algorithm. The results show a significant improvement over the Otsu's method in terms shadow removal, document legibility and optical character recognition. The results also compare well with the state of the art Sauvola's method results.