A dynamic window based method for binarization of document images

Thresholding is an important step in the identification of text in document images. A no. of global and local thresholding techniques have been proposed for the binarization of document images. In this paper we present a a new dynamic window based thresholding approach where the thresholding is performed locally over a dynamic window by means of a hard clustering based approach. A performance comparison of the proposed method is also carried out with some of the widely used thresholding techniques.