Robust Locally Weighted Regression for Superresolution Enhancement of Multi-Angle Remote Sensing Imagery

This paper presents a robust locally weighted least-squares kernel regression method for superresolution (SR) enhancement of multi-angle remote sensing imagery. The method is based on the concept of kernel-based regression, where the local image patch is approximated by an \mbiN-term Taylor series. To reduce the impact of high frequency noise on SR performance, a robust fitting procedure is adopted. The approach proposed is tested with simulated multi-angle data derived from panchromatic WorldView-2 imagery and with real multi-angle WorldView-2 remote sensing images.