Image segmentation using region derived triple thresholding

Abstract

In image processing, segmentation is the process of partitioning an image on the basis of intensity values of pixels, which makes it easier to analyze the image and distinguish objects and boundaries. The paper puts forward a method that clusters pixels into four regions based on their intensities. The method uses the process of thresholding with two local thresholds and one global threshold. The novelty of this process lies in the automated generation of three thresholds based on intrinsic characteristics of the image. The global threshold is obtained through neighborhood comparison of localized regions spanning through the image. The global threshold forms two sets of pixels, from which the local thresholds are obtained.