SAR images matching based on local shape descriptors

For SAR Images Navigation images matching according to keypoints and feature descriptors is a key technology. Firstly, the novel algorithm detects local extrema in Zoser Pyramid and assigns their orientations. Secondly, it extracts edges by Canny Detector and for each keypoint tests its feature vectors with 49 dimensions by statistic histograms method according to relative distances and orientations between the keypoint and other surrounding points on edges. Lastly, it gets corresponding pixels of two images by matching between two Descriptors. The algorithm has matching in variance to image displacement, scale and rotation, and is shown to provide robust matching across a substantial range of affine distortion, change in 3D viewpoint, addition of noise, and change in illumination. Because SAR images are often blurred and lack of stable details, the algorithm can achieve more reliable recognition and process 3 times quicker than SIFT.