An Algorithm for Histogram Median Thresholding

Abstract

One algorithm for thresholding using a histogram median is presented in this paper. The algorithm is named HisMedian and it is implemented on Java. It is also proposed a taxonomy of thresholding algorithms based on a method for defining threshold value as a real color of image or calculated color's value. According to the proposed taxonomy a set of popular thresholding algorithms (including HisMedian) is experimentally evaluated using three test images. The experimental results show that if a histogram is bimodal then the algorithms which use real color(s) from the image as a threshold(s) achieve better results than algorithms which use calculated value(s) as a threshold(s).