Abstract: The high noise and local deformation of multi-modal images reduce the accuracy of scale invariant feature transform (SIFT) image matching. To solve this problem, a new method based on the ...