Unsupervised change detection in multitemporal SAR images via NSCT-domain feature clustering

In this paper, a new unsupervised multitemporal change detection method for synthetic aperture radar (SAR) imagery is proposed. This method follows the “comparison-classification” framework where the difference image is obtained by comparing the multitemporal SAR images acquired on the same geographical area at two different time instances and the change map is generated by classifying the difference image into “changed class” and “unchanged class” using feature clustering technique. The features that are clustered result from the multiscale and multiband decomposition of the difference image with the nonsubsampled contourlet transform (NSCT). Experimental results on a pair of synthetic multitemporal SAR images and a pair of real multitemporal SAR data justify the effectiveness of the proposed method. Comparison with the UDWT-based method is also given, showing better detection performance of the proposed method.