Half-quadratic Based Robust Sparse Hyperspectral Unmixing Framework

We present a general half-quadratic based hyperspectral unmixing (HU) framework to solve the robust or sparse unmixing problem. A series of potential methods can be designed and developed to solve HU problem through this framework. By introducing correntropy metric, a correntropy based spatial-spectral robust sparsity regularized (CSsRS-NMF) unmixing method is derived through the proposed framework to achieve two-dimensional robustness and adaptive weighted sparsity constraint for abundances simultaneously.