Scale and rotation robust line-based matching for high resolution images

A line-based matching method is proposed to overcome the low significant level of point feature and the shortage in the matching between weak texture images. Edges are extracted to generate line segments at first. Then, saliency-lines are detected and a control network is constructed in each image. Every remaining general-line is grouped into a saliency-line through a process of clustering. Finally, an iterative algorithm is applied to search for correspondences based on their position relative to the saliency-line. Sufficient spatial information is available to reduce the ambiguity and avoid false matches. Experimental results demonstrate that the proposed method is robust to image scale change and rotation, and the performance is better than point-based method in low texture area.