Automated registration of synthetic aperture radar imagery to LIDAR

The registration of synthetic aperture radar imagery to other images is difficult, especially in mountainous terrain. We introduce a new approach to this problem that registers radar images to digital elevation models derived from LIDAR. The algorithm generates a predicted image from the elevation model using the radar geometry and then registers the predicted image to the radar image with normalized cross correlation. The algorithm is automated, fast, and accurate. It georegisters a large satellite image in only a few minutes on a PC and with an absolute geospatial error of less than two meters. In addition to georegistration, the algorithm is useful for image registration and the fusion of radar images to multispectral images. Results on satellite imagery are presented along with accuracy measurements and cross-sensor registration results.