The Research On Denoising Model And Algorithm Of Hyperspectral Image Based On Total Variation And Low-Rank Representation

Hyperspectral images（HSIs）are always contaminated by various mixed noise（e.g.Gaussian noise,impulse noise,stripe noise,and deadlines）,which degrades the quality of acquired images and seriously affects the subsequent extensive applications.Therefore,hyperspectral image denoising is of great significance for obtaining rich image information and subsequent application.Total variation（TV）is popular for its capability of preserving edge information and promoting smoothness in HSI denoising.However,TV may lead to a series of problems such as over-smoothness and loss of detail in denoised images.To tackle the above prob-lems,we propose a weighted double sparsity total variation and low-rank representation denoising model（LRDSTV）for the mixed noise removal.Specifically,the double spar-sity means fiber sparsity with sparse fibers in the gradient domain,and in this paper,we also adopt the weighted strategy to promote the sparsity of difference images.Moreover,considering the spectral correlation of HSIs,we utilize the weighted nu-clear norm to explore the low-rank property of mode-3 unfolding of the HSI.Then,the alternating direction method of multipliers（ADMM）is applied for the optimization of the LRDSTV model.We prove the relative and objective convergence of the proposed algorithm.Finally,a series of denoising experiments on simulated and real data sets are showed in this paper,and we compare with some state-of-the-art algorithms in multiple as-pects,such as visual comparison,quantitative comparison,qualitative comparison,which demonstrates the relative effectiveness and superiority of the new algorithm.