A Bayesian approach to unsupervised multiscale change detection in synthetic aperture radar images

In this paper, an unsupervised change detection technique for synthetic aperture radar (SAR) images acquired on the same geographical area but at different time instances is proposed by conducting probabilistic Bayesian inferencing with expectation maximization-based parameter estimation to perform unsupervised thresholding over the data collected from the dual-tree complex wavelet transform (DT-CWT) subbands generated at the various scales. The proposed approach exploits a DT-CWT-based multiscale decomposition of the log-ratio image, which is obtained by taking the logarithm of the pixel ratio of two SAR images, aimed at achieving different scales of representation of the change signal. Intra- and inter-scale data fusion is performed to enhance the change detection performance. Experimental results obtained on SAR images acquired by the ERS-1, and JERS satellites confirm the effectiveness of the proposed approach.