SAR Images Change Detection Based on Comparison of Two-dimensional Probability Density Functions

In this paper, the tradition change detection method based on local statistical feature is expanded to two-dimensional feature space, and a SAR image change detection method based on comparison of two-dimensional probability density functions is proposed. In this method, the values of adjacent pixels are combined to build two-dimensional observation vector. Then, in each temporal image, the Probability Density Function (PDF) of the vector is estimated by two-dimensional Gram-Charlier expansion. On the basis, change detection is fulfilled by computing the K-L divergence between the PDFs in different temporal images. Experiment results show that the proposed algorithm has better performance than the traditional method.