A Novel Visual Saliency Guided High Resolution Remote Sensing Image Change Detection Algorithm

This paper presents a novel visual saliency guided high resolution remote sensing image change detection (CD) algorithm. This approach is based on closed-loop process made up of the three procedures. (1) Generating the difference image as the foundation for further processing; (2) Use the modified saliency extraction algorithm to select the regions with higher attractiveness; (3) Adopt the revised fuzzy clustering to finalize the real changed regions (RCR). Our algorithm could fuse the traditional change analysis patterns with the saliency prior which could enhance the overall efficiency. Experiments carried out on the different sets of high resolution multi-temporal images acquired by the satellite. The experimental result shows that our method could achieve better performance.