A multi-source image registration algorithm based on combined line and point features

A multi-source image registration algorithm based on combined line and point features is proposed for images containing typical line objects. Firstly, the image control line features are extracted for coarse registration by the use of visual saliency and Line Segment Detection (LSD). Visual saliency represents human visual characteristics. LSD has attributes including rotation invariance, illumination changes insensitivity and noise resistant ability. Secondly, Scale Invariant Feature Transform (SIFT) based on multi-resolution analysis is used to extract the point features with scale and rotation invariant characteristics. Then the feature points are used to realize the fine registration. Finally, the simulation results are analyzed, and the validity of the algorithm is verified from subjective effect and objective evaluation indices.