Hyperspectral Image Denoising Based on Spatial Total-Variation Group Sparse and Multilayer Low-Rank Prior

针对高光谱图像的噪声去除问题，本文提出了一种基于空间变分组稀疏和多层空间–光谱先验的高光谱图像去噪方法。在该方法中，我们使用张量表示高光谱图像，并对其进行张量张量积分解。为了进一步提升去噪效果，我们提出了空间变分组稀疏，用于更好地捕捉高光谱图像的稀疏结构；其次，我们对张量张量积分解得到的系数张量继续进行张量张量积分解，并对分解后的两个张量施加加权Schatten-p范数约束，以增强对图像稀疏特征的建模能力，最后使用交替方向乘子法对去噪模型进行求解。实验结果表明，本文所提出的去噪算法在性能上优于其他对比算法，验证了该方法的先进性。To address the problem of noise removal from hyperspectral images, this paper proposes a method for denoising hyperspectral images based on spatial variable group sparsity and multilayer spatial-spectral prior. In this method, we use a tensor to represent the hyperspectral image and decompose it into a tensor tensor product. To further enhance the denoising effect, we propose spatial variable group sparsity for better capturing the sparse structure of the hyperspectral image; second, we continue the tensor tensor product decomposition on the coefficient tensor obtained from the tensor tensor product decomposition, and impose a weighted Schatten-p paradigm constraint on the decomposed two tensors to enhance the ability to model the sparse features of the image, and finally, we use the alternating direction multiplier method to solve the denoising model. The experimental results show that the denoising algorithm proposed in this paper outperforms other comparative algorithms in terms of performance, which verifies the advancement of the method.