Unsupervised Change Detection Based on Iterative Histogram Matching and Bayesian Decision of Thresholding

Abstract

This paper presents a novel unsupervised change detection (CD) algorithm for remote sensing images based on iterative histogram matching and bayesian decision of thersholding. In each iteration, histogram matching (HM) map is improved, leading to more accurate radiometric calibration. Then, difference image is generated from the change detection method-image differencing for further analysis, in which an unsupervised threshold selection algorithm based on Bayesian decision theory is used, aiming at extracting the changed information automatically from the images. This algorithm is suitable for those images which have great radiometric difference as well as great changes from each other. The experimental result of the proposed algorithm compared with the traditional way of CD is presented, which indicate that the proposed method improves the result effectively and is superior to the traditional one.