Satellite Image Change Detection Using Laplacian–Gaussian Distributions

In this paper, Bayesian based change detection (CD) algorithm is proposed for satellite images, using prevalent probability distributions in spatial domain. The log ratio difference image (DI) is generated from two images acquired over the same area at different time instants. DI is clustered into two classes namely changed and unchanged using K-means clustering for basic discrimination. For further progress, changed pixels are modeled with Laplacian probability density function (pdf) whereas, unchanged pixels with Gaussian pdf. Expressions are derived for the parameters of pdfs using iterative free negative order fractional moments in order to avoid convergence problems. Binary CD map is generated using Bayesian threshold. Experimental results show that the proposed method is effective by achieving better performance in terms of false alarm as 2.3% and overall accuracy as 97.7% in less computing time compared with the state-of-the-art methods.