Graph-regularized tensor robust principal component analysis for hyperspectral image denoising.
Graph-regularized tensor robust principal component analysis for hyperspectral image denoising.
复制标题
DOI：
10.1364/ao.56.006094
复制
发表时间：
2017-08
期刊：
影响因子：
1.9
通讯作者：
Yongming Nie;Linsen Chen;Hao Zhu;S. Du;Tao Yue;Xun Cao
中科院分区：
文献类型：
--
作者：
Yongming Nie;Linsen Chen;Hao Zhu;S. Du;Tao Yue;Xun Cao
本文提出了一种新的高光谱图像去噪模型--图正则张量稳健主成分分析(GTRPCA)。将谱图正则化引入到TRPCA中，通过保留嵌入在高维空间中的局部几何结构，使模型更加准确。在张量奇异值分解(t-SVD)的基础上，提出了一种基于张量变换方向的乘子变换(ADMM)算法。在合成数据集和实际捕获数据集上的实验证明了该方法的有效性。
In this paper, we have developed a novel model that is named graph-regularized tensor robust principal component analysis (GTRPCA) for denoising hyperspectral images (HSIs). Incorporating spectral graph regularization into TRPCA makes the model more accurate by preserving local geometric structures embedded in a high-dimensional space. Based on tensor singular value decomposition (t-SVD), we introduce a general tensor-based altering direction method of multipliers (ADMM) algorithm which can solve the proposed model for denoising HSIs. Experiments on both the synthetic and real captured datasets have demonstrated the effectiveness of the proposed method.