Performance Evaluation of Thresholding Techniques on Modi Script

Abstract

Images are degraded during image procurement in terms of several disturbances. As a result, poor illumination, low contrast value, background blurring etc. occur in images that hampers its quality. Using thresholding process, we attempt to improve eminence of image, enhance contrast value as well as separates foreground and background information. There exist several prevailing thresholding techniques such as bernsen, wolf, sauvola, otsu, niblack and bradley for numerous scripts. A threshold model is proposed in this paper for binarizing Modi character images efficiently. It uses various global and local thresholding techniques in order to achieve high contrast value, better illumination etc. Two performance parameters such as mean square error and peak signal to noise ratio is used to measure the effects of varied thresholding techniques. As a result, Otsu thresholding technique shows effective outcome for binarizing Modi vowels in more appropriate form.