Cui, SC; Yang, SZ; Qiao, YL; Zhao, Q; Wang, JC; Wang, Z (2012). Adaptive regularized filtering for BRDF model inversion and land surface albedo retrieval based on spectrum cutoff technique. OPTIK, 123(3), 250-256.
A regularized inversion method is presented to invert the bidirectional reflectance distribution function (BRDF) model for albedo calculations. It is known as discrete ill-posed inverse problem. Essential issues that must be considered include solution existence, solution uniqueness, instability of the solution process, and geophysical meanings of solution. In order to robustly estimate BRDF models parameters, a spectrum cutoff technique is employed to suppress spurious oscillations due to poor angular distribution, sparse sample size, or both. We studied difficulties in estimating BRDF parameters and analyzed the cause of the instability in the solution, and then presented a novel algorithm. The proposed RSC (Regularized Spectrum Cutoff) method is based on improving the spectrum of the BRDF kernel matrix and can always find suitable BRDF parameters for sparse satellite observations. The algorithm is compared with the MODIS (Moderate Resolution Imaging Spectroradiometer) full retrieval (FR) algorithm, and it is tested to be suitable for the case of sufficient observations. Numerical experiments with POLDER (Polarization and Directionality of the Earth's Reflectances) BRDF datasets and MODIS images suggest that the developed spectrum cutoff method is adaptive and robust, and it is valuable for retrieving land surface BRDF and albedo. (C) 2011 Elsevier GmbH. All rights reserved.