Hyperspectral Image Restoration for Non-additive Noise

This work is about a curious phenomenon. Among hyperspectral image (HSI) restoration methods, most of them suppose the observed HSI is the superposition of clean component and noise component. However, in real world applications, the components are often non-additive. The non-additivity admits the uniqueness of attribute. Especially each pixel of image belongs to either clean component or noise component. To separate components robustly, this work proposes a general HSI restoration framework. The novelty of this work lies in exploiting non-additivity of components by a binary mask matrix. Two special cases are discussed. To solve the proposed models, efficient alternating minimization algorithms are developed. Compared with gradient-type algorithms, our algorithms are easy to implement as there is no need to tune optimization parameters like step sizes. The experimental results over simulated and real datasets verified the performance of the proposed method.