Edge Enhanced Optimum Automatic Thresholding

Abstract

The thresholding process has been widely used to segment one or several objects from their background. The automatic thresholding requires that, the image formation be of two well-segmented homogeneous regions, and there exists a threshold value that separates these regions. The separation deteriorates when the objects are not homogenously illuminated. In this paper a new algorithm is proposed for the optimum automatic thresholding of digital images in the presence of non-homogenous illumination. The algorithm is based on determining an optimum threshold value that maximizes an edge similarity function of the binary and the original gray level images. Experimental results to compare the proposed algorithm to the various thresholding techniques are also presented.