I. Introduction
Hyperspectral image (HSI) with hundreds of spectral channels can reflect wealthy spatial information of real scenes and facilitate many applications in geography, agriculture, and military [1]. However, during the imaging process, real-world HSIs are often contaminated by various noises, such as Gaussian noise, impulse noise, and stripes [2], [3], which restricts subsequent applications and analysis. Therefore, restoring clean HSI plays a critical role in HSI-related tasks.