Registration and BRDF extraction of airborne multi-angular images in different resolutions

An auto-registration algorithm is designed for multi-angular remote sensing, which refines the image warping function in a step-by-step way. It first uses a wavelet transform to degrade the original images to a series of low resolution ones, then constructs warping function with B-splines, and refines the B-splines according to control points found by a correlation based method. The algorithm is tested with images acquired in North-China by an airborne multi-angular TIR/VNIR imaging system. Results show the advantage of B-splines in comparison with polynomial or thin-plate splines. After registration is done, we give an example for BRDF extraction.