Hyperspectral image denoising via weighted double sparsity total variation and low-rank representation (Q6593068)

Hyperspectral images (HSI) store simultaneously hundreds of continuous frequency (i.e. color) bands and are contaminated by various noises (Gaussian noise, impulse noise, stripes, deadlines), which degrades the image quality. Introduced in [\textit{Q. Yuan} et al., IEEE Trans. Geosci. Remote Sensing 50, No. 10, 3660--3677 (2021; \url{doi:10.1109/TGRS.2012.2185054})], the denoising technique -- total variation (TV) in which the spectral noise differences and spatial information differences are both considered in the process of noise reduction -- can preserve details and promote smoothness, but may also cause over-smoothness and loss of details. Here, the authors propose a double sparsity TV and low-rank representation denoising model which uses the fiber sparsity in the differential image of the HSI and also considers the sparsity of individual fibers. In this paper, two denoising algorithms based on the alternating direction method of multipliers are described and the effectiveness of the algorithms is demonstrated with experiments on simulated and real hyperspectral images.