A fully automatic registration approach based on contour and SIFT for H J-1 images

To achieve a fully automatic registration between HJ-1 CCD images and HJ-1 infrared images is a difficult task as it must deal with the varying illuminations and resolutions of the images, different perspectives, and the local deformations within the images. In this paper, aimed at those registration issues, a fully automatic registration approach based on contour and SIFT is proposed. The registration technique performs a pre-registration process using contour feature matching algorithm that decides the overlapping region between a reference image and an input image. Once the coarse regions are obtained, it performs a fine registration process based on SIFT detector and a local adaptive matching strategy. In the fine registration process, image blocking theory is used, which not only speeds up the features extraction and matching, but also makes the matching point pairs distributed uniformly in images, and further improves the accuracy of input image rectification. Experiments with visible images and infrared images from HJ-1A/B demonstrate the efficiency and the accuracy of the proposed technique for multi- source remote sensing images registration.