Change detection in SAR images by means of grouping connected regions using clone selection algorithm

Most of the change detection techniques for SAR images based on the analysis of the difference image always generate the result map at the pixel level, which leads to a noisy change detection map, with holes in connected components and jagged boundaries. To attack this problem, a novel technique to generate the result at the region level is proposed, which considers each connected region as an agent and then gives the optimal label combination for all agents through the clone selection algorithm and a new fitness function. The region oriented method can avoid the above problem and improve the change detection performance. Experimental results on the real Radarsat SAR dataset validate the effectiveness of the method on both the quantitative and subjective aspects.