Local Low-Rank and Sparse Representation for Hyperspectral Image Denoising

Hyperspectral image (HSI) denoising is a fundamental task in a plethora of HSI applications. Global low-rank property is widely adopted to exploit the spectral-spatial information of HSIs, providing satisfactory denoising results. In this paper, instead of adopting the global low-rank property, we propose to adopt a local low rankness for HSI denoising. We develop an HSI denoising method via local low-rank and sparse representation, under an alternative minimization framework. In addition, the weighted nuclear norm is used to enhance the sparsity on singular values. The experiments on widely used hyperspectral datasets demonstrate that the proposed method outperforms several state-of-the-art methods visually and quantitatively.