Land-Cover Change Detection for SAR Images Based on Biobjective Fuzzy Local Information Clustering Method With Decomposition

The existence of a speckle noise significantly affects the accuracy of land-cover change detection results for synthetic aperture radar (SAR) images. This letter proposes a biobjective fuzzy local information clustering method with decomposition (BIFLICM/D) to address this problem. SAR images change detection is described as a biobjective fuzzy local information clustering problem from the aspects of preserving image details and removing noise. To improve the ability to extract the original information detail, the log-mean ratio method is used to generate a first difference image in BIFLICM/D. The second difference image is achieved by combining the homomorphic filtering and saliency detection, which effectively removes the speckle noise. Fuzzy clustering objective functions are then constructed for the two difference images to recognize the changed and unchanged pixels under different requirements. A new fuzzy membership degree-updating method is adopted to optimize the two objective functions, which can balance the influences of the two objectives and improve the robustness of the proposed method. The experimental result demonstrates that the proposed method is more effective and superior to the comparison algorithms.