Multi-thresholds Fusion Based Change Detection of SAR Images by Combining the Intensity and Coherence Information

In this paper, we investigate the change detection techniques for SAR images by using the statistical distribution models of the intensity and coherence information with a multi-thresholds fusion method. According to the statistical characteristics of the intensity and coherence information of SAR images, we first select suitable distributions to model the intensity and coherence information respectively, and use Jeffrey distance as difference metric to obtain difference maps. We then employ three different thresholding segmentation methods to obtain multiple initial results of change detection. Finally, we use Markov Random Field to fuse multiple change detection maps for achieving the final result. Experiments show that our fusion strategy can obtain more stable results compared with the optimal threshold based methods.