Automatic satellite image registration by combination of matching and random sample consensus

We propose a new algorithm for automated registration of satellite images. Registration means finding the relation between image coordinates and a reference coordinate system. The algorithm consists of two steps. The first one is the automated generation of control points. An automated matching based on normalized cross correlation is used. We have improved the accuracy of matching by determining the size and shape of match windows according to incidence and scene orientation angles. The second one is the robust estimation of mapping functions from control points. We used the random sample consensus (RANSAC) algorithm for this step. We argue that it is the second step which gives the robustness of any automated registration algorithms. We carried out experiments with SPOT images over three test sites. Through matching, a number of control points were generated. RANSAC was applied to the control points. All outliers were correctly identified for all three test sites and mapping functions estimated without outliers. The accuracy of estimation was comparable to that of estimation with control points generated all by manual measurements. The results support that our algorithm can be used for robust automated registration.