Hyperspectral image denoising with rare signal preserving by jointly filtering image component

Recently a multiway Wiener filter (MWF) has been proposed to overcome the problems in the hyperspectral image (HSI) denoising. Though MWF performs better than the classical methods, it has weak capability in rare signal preserving due to its strategy of treating the HSI as a whole entity. In this paper, a jointly component filtering way is discussed to improve the performances of MWF in the scenes with rare signals. In the jointly component filtering algorithm, the dataset is no longer treated as an entire entity and instead the components extracted from the data are filtered jointly in each mode. The rare signal preserving ability of the algorithm is proven and experiments on both simulated and real-world datasets show that the jointly component filtering performs better than MWF in rare signal preservation.