Automatic Registration Of SAR And Optical Image Based On Line And Graph Spectral Theory

Abstract. In this paper, a novel registration method is proposed by integrating the graph spectral theory and line features. The principal steps of our algorithm are as follows. Firstly, the images are filtered to enhance the reliability and robustness of registration, and line features are acquired by Hough Transform. Secondly, the original point features can be obtained by calculating the line intersections. The points are normalized to reduce computational complexity. Thirdly, voronoi diagrams of two point sets are extracted respectively. The original corresponding point sets are determined by corresponding voronoi diagrams, which can be obtained by Graph Spectral Theory. At last, RANSAC is used to remove the wrong corresponding points. The transform relationship of the two input images can be achieved using the corresponding point sets. The experimental results show that the proposed method can achieve high accuracy for the registration between optical and SAR images.