Automatic Remote-Sensing Images Registration by Matching Close-Regions

Abstract

Remote-sensing images registration is a fundamental task in image processing, which is concerned with establishment of correspondence between two or more pictures taken, for example, at different times, from different sensors, or from different viewpoints. Because of the different gray level characters in such remote-sensing images, it’s difficult to match them automatically. We usually constrain the images to some particular categories, or do the job manually. In this paper, we develop a new algorithm for remote-sensing images registration, which takes full advantage of the shape information of the close-regions bounded by contours after detecting and linking the edges in images. Based on the shape-specific points of the close-regions, we match the close-regions by evaluating their matching degrees. Using the matched pairs of the close-regions, the geometric parameters for images registration are computed and this registration task can be performed automatically and accurately. This new algorithm works well for those images where the contour information is well preserved, such as the optical images from LANDSAT and SPOT satellites. Experiments verified our algorithm, and showed that the performance of executing it sequentially depends a lot on the size of the input images. The time complexity will increase exponentially as the size of images increases. So we extend the sequential algorithm to a distributed scheme and perform the registration task more efficiently.KeywordsFeature PointInput ImageReference ImageImage RegistrationMatched PairThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.