Robust Graph Regularized Non-negative Matrix Factorization for Image Clustering
Robust Graph Regularized Non-negative Matrix Factorization for Image Clustering
复制标题
DOI：
10.1007/978-3-030-64221-1_21
复制
发表时间：
2020
期刊：
影响因子：
--
通讯作者：
Xiangguang Dai;Keke Zhang;Juntang Li;Jiang Xiong;Nian Zhang
中科院分区：
文献类型：
--
作者：
Xiangguang Dai;Keke Zhang;Juntang Li;Jiang Xiong;Nian Zhang
非负矩阵分解及其变体已被用于计算机视觉和机器学习，然而，当数据集被离群值和噪声破坏时，它们无法实现鲁棒的分解。本文提出了一种鲁斯特图正则化非负矩阵分解方法（RGRNMF）用于图像聚类。为了提高被离群点和噪声污染的图像数据集的聚类效果，我们提出了一个加权约束的噪声矩阵，并施加流形学习到低维表示。实验结果表明，RGRNMF算法在受椒盐噪声和连续遮挡污染的人脸数据集上具有较好的聚类性能。
Non-negative matrix factorization and its variants have been utilized for computer vision and machine learning, however, they fail to achieve robust factorization when the dataset is corrupted by outliers and noise. In this paper, we propose a roust graph regularized non-negative matrix factorization method (RGRNMF) for image clustering. To improve the clustering effect on the image dataset contaminated by outliers and noise, we propose a weighted constraint on the noise matrix and impose manifold learning into the low-dimensional representation. Experimental results demonstrate that RGRNMF can achieve better clustering performances on the face dataset corrupted by Salt and Pepper noise and Contiguous Occlusion.