A Fast Registration Method Based on Line Features

Abstract

A fast registration method based on line features is proposed for optical remote sensing images with typical line features. Firstly, the line features are extracted by Line Segment Detection (LSD). LSD has attributes including rotation invariance, illumination changes insensitivity and noise resistant ability. Secondly, the initial fast image registration is implemented based on features consensus, to eliminate the influence of scaling, rotation and translation between images on the matching relationship of the lines. Then the line segments with same name are determined according to the proposed similarity criterion and the bidirectional matching relation. By utilizing the intersection points of these line pairs with same name as the matching control points, fine registration can be realized. Finally, the simulation results are analyzed, accuracy and computational efficiency of the proposed method are verified from subjective effect and objective evaluation indices.