Line Based SAR and Optical Image Automatic Registration Method

In SAR and optical images, invariant feature extraction and matching is very difficult, linear feature extraction is relatively easy. This paper presents a linear feature based image registration method, which using the distance of line segment´s endpoints to another line segment and the distance between the two line segments´ centers defines a spatial coherence measure between two line segments, the image registration problem is transformed into Calculate the maximum of spatial coherence measure between line segments, in order to obtain the registration parameters we use the genetic algorithm to search the global maximum. Since the correct matched line segments has a great contribution in the objective function, isolated line segments are inhibited very well, the robustness of this method is good, a small number of line segments can also converge to the correct registration parameters. Experimental results show that the image registration method is feasible and effective, pixel level accuracy can be achieved.