Abstract: To resolve multi-sensor remote sensing image registration, an algorithm combining scale invariant feature transform(SIFT) point features and Canny edge feature matching was proposed. First,this algorithm uses SIFT to extract point feature and conduct image coarse registration. Then after getting the initial affine transformation parameters,this algorithm uses Canny algorithm to extract edge feature and then uses the cost function for edge matching. Last, after getting the effective matching feature points by filtering outliers, this algorithm conducts image accurate registration. I}his algorithm combines the advantages of SIFT and Canny, and solves the problem that the multi-sensor remote images are difficult for correct registration because of geometric differences and radiometric differences. Experimental results show that this algorithm has strong robustness, and achieves high registration accuracy.

Key words: Multi-sensor image, Image registration, Feature matching, SIFT, Canny edge detection