A Change Detection Method for Man-Made Objects in SAR Images Based on Curvelet and Level Set

Abstract

An unsupervised change detection method for man-made objects in co registered multi-temporal SAR images is proposed in this paper. Based on analyzing the edge structure property of man-made objects, the Curve let transform is used to denoise and enhance the difference image by manipulating certain Curve let coefficients. Then, the enhanced difference image is segmented into the changed and unchanged regions by level set method. Some prior knowledge of man-made objects in SAR images is exploited in both steps. The proposed method can overcome the drawbacks of traditional pixel-level change detection methods, and obtain robust detection results even for high level speckle noise. Experimental results demonstrate its effectiveness and feasibility.