The invention discloses a method for matching images of different sources based on a fit contour. The method is characterized by comprising the following steps of: first, in order to overcome noises, such as transformation, mutation, burr and the like, of metric space, performing curvature scale space-based contour denoising and fitting processing on extracted contours to obtain smooth contours on the same scale; then, performing centroid distance-based contour description on closed contours, and performing local curvature-based contour description on contours formed by open curves; next, establishing RST (rotating, scaling and translating) feature invariants on the contour description; and finally, performing corresponding reconstruction on the feature invariants to obtain a final matching point pair. By the method, the matching of the images of the different sources based on the contour is realized; and good robustness on the matching of the images of the different sources with rotating, translating and scaling changes is achieved.