Automatic Image Orientation

Our research team has developed a system for fully automatic calibration and orientation of image sequences exploiting sparse matching techniques and closed form algorithms. In a first step, image pairs are identified from unordered sets of – terrestrial or aerial (UAVs) – images. Sparse matching is then performed employing state-of-the-art features with their des­criptors (such as SIFT and SURF) at multiple image scales. In this way image matches are established across different stereo-pairs leading to multi-image point correspondences. By means of closed form algorithms, initial image orientations are computed and finally exterior and interior orientation parameters are refined through a typical bundle adjustment solution.