In the process of mapping using optical remote sensing images

due to the limit of the conditions obtaining the remote sensing data and the affect of weather

the obtained images often are covered by clouds

which causes the losing of the information in local area and affects the quality of image map in some degree. In this paper

a method is proposed by fusion of multi sensor images collected at different date to remove the affect due to cloud cover. It consists in two steps: firstly the auto registration is done around the region with clouds. In this step area based grey match is used to get tie point

and statistical relational matching is done iterative to adjust the initial match result more reliable. So the two image can be registrated local area with high accuracy according to the final matching points. Secondly the image with cloud is replaced automatically by the other sensor image which has been radiation corrected through histogram matching so as to prevent the false edge from emerging during image mosaicing. At last

the test that SPOT image with clouds is replaced by TM image is done. The result demonstrates that the proposed method is possible and valuable

and the information in cloud area is recovered properly.