基于加权块稀疏联合非凸低秩约束的高光谱图像去条带方法

Hyperspectral images destriping approach with weighted block sparsity regularization and non-convex low-rank penalty

针对高光谱图像(hyperspectral images,HSI)去条带易引起影像结构细节丢失问题,提出一种基于加权块稀疏(weighted block sparsity,WBS)正则化联合最小最大非凸惩罚(minimax concave penalty,MCP)约束的HSI去条带方法。本算法采用加权ℓ_(2),1范数和MCP范数对条带稀疏结构和低秩约束,ℓ_(1)范数对干净图像结构保持正则化约束,构建加权块稀疏和MCP约束的条带去除模型,采用交替方向乘子(alternating direction method of multipliers,ADMM)算法迭代求解对应模型,重建获得干净的HSI图像。实验结果表明,提出方法在实际HSI的平均等效视数从28.45提高到83.47,边缘保持指数较其他算法至少增加0.056,特别是对于非周期条带噪声,采用自适应权值更新稀疏水平,增强了组稀疏性,在保持影像边缘和加强区域平滑性方面性能更佳,去噪声效果更好。

To solve the loss of image details when removing the stripe noise of hyperspectral images(HSI),a destriping method for hyperspectral images was proposed,which was based on the weighted block sparsity(WBS)regularization and the minimax concave penalty(MCP)constraint(WBS-MCP).WBS-MCP was constructed by employing weightedℓ_(2),1 norm and MCP norm for the sparse features and low-rank constraints,and applyingℓ_(1) norm to the edge smoothing constraint of clean image.The alternating direction method of multipliers(ADMM)algorithm was used to iteratively solve the corresponding constraint model,and the clean HSI image was obtained by reconstruction.The experimental results show that the mean equivalent number of look of real HSI improves from 28.45 to 83.47,and the edge retention index increased by 0.056 at least,especially for the aperiodic stripe noise.The adaptive weight was used to update the sparse level,which could enhance the group sparsity and has better performance in maintaining the image edge and enhancing the area smoothness,so that the noise removal has better effect on aperiodic stripe noise.