Change detection in unregistered optical satellite images using combinatorial clustering method

In this paper, we propose a new change detection method for unregistered multitemporal satellite images. This method is able to obtain the changed information by using the combinatorial clustering method. To obtain the change detection map, firstly, the method of oversegmentation is employed for the input images to reduce the computational cost. Secondly, we apply the combinatorial clustering method, which integrates the methods of bottom-up unsupervised clustering and the top-down discriminative clustering, to obtain the segmentation score for the result of over segmentation. Finally, the change detection map is obtained by thresholding the score. Experimental evaluation on the multitemporal satellite images demonstrates that the proposed method has good performance on the unregistered images.