Polarimetric Inverse Scattering via Incremental Sparse Bayesian Multitask Learning

摘要

In this letter, we employ the sparse Bayesian multitask learning to realize joint sparsity-enforcing polarimetric inverse scattering. The prior assumption about the data model is redesigned to avoid information sharing across unrelated tasks. Based on this assumption, we provide the formulas for Bayesian inference as well as the algorithm flowchart, which still has the linear complexity. Experimental results demonstrate that polarimetric inverse scattering with the proposed method can effectively extract the characteristics of canonical scatterers.