Robust Low-rank subspace segmentation with finite mixture noise

Robust Low-rank subspace segmentation with finite mixture noise
复制标题

具有有限混合噪声的鲁棒低秩子空间分割

DOI:
10.1016/j.patcog.2019.03.028
复制
发表时间:
2019-09
影响因子:
8
通讯作者:
Wang Jun
Wang Jun
中科院分区:
计算机科学1区
文献类型:
--
作者:
Guo Xianglin;Xie Xingyu;Liu Guangcan;Wei Mingqiang;Wang Jun

文献摘要

参考文献

被引文献

相似文献

当处理高维数据中存在的复杂噪声时,子空间分割或聚类仍然是计算机视觉中感兴趣的挑战。目前的大多数稀疏表示或最小秩算法都是建立在ℓ1-范数或ℓ2-范数损失的基础上的,这对离群点很敏感。有限混合模型作为一类建模复杂噪声的强大而灵活的工具,成为一种必不可少的工具。在所有的选择中,指数族混合是非常有用的,因为它对任何连续分布都具有普遍的逼近能力,因此涵盖了更广泛的噪声分布特征。基于这样的建模思想,本文重点研究了基于有限混合指数幂分布的复杂噪声污染的子空间聚类问题。然后,我们利用惩罚似然函数来执行自动模型选择,从而避免过度拟合。此外,我们还对表示矩阵的奇异值引入了一种新的先验知识,从而在我们的非凸非光滑优化中得到了一种新的惩罚。MoEP模型的参数可以用最大后验概率(MAP)方法估计。同时,利用联合加权ℓp-范数和Schatten-q拟范数极小化算法计算子空间。理论和实验结果都证明了该方法的有效性。
Subspace segmentation or clustering remains a challenge of interest in computer vision when handling complex noise existing in high-dimensional data. Most of the current sparse representation or minimum-rank based techniques are constructed on ℓ1-norm or ℓ2-norm losses, which is sensitive to outliers. Finite mixture model, as a class of powerful and flexible tools for modeling complex noise, becomes a must. Among all the choices, exponential family mixture is extremely useful in practice due to its universal approximation ability for any continuous distribution and hence covers a broader scope of characteristics of noise distribution. Equipped with such a modeling idea, this paper focuses on the complex noise contaminated subspace clustering problem by using finite mixture of exponential power (MoEP) distributions. We then harness a penalized likelihood function to perform automatic model selection and hence avoid over-fitting. Moreover, we introduce a novel prior on the singular values of representation matrix, which leads to a novel penalty in our nonconvex and nonsmooth optimization. The parameters of the MoEP model can be estimated with a Maximum A Posteriori (MAP) method. Meanwhile, the subspace is computed with joint weighted ℓp-norm and Schatten-qquasi-norm minimization. Both theoretical and experimental results show the effectiveness of our method.
DOI: 10.1109/tnnls.2014.2377477
发表时间: 2015-02
影响因子: 10.4
作者:
Jinhui Chen;Jian Yang;Lei Luo;J. Qian;W. Xu
通讯作者: Jinhui Chen;Jian Yang;Lei Luo;J. Qian;W. Xu
DOI: 10.1007/bf01581204
发表时间: 1992-07-06
影响因子: 2.7
作者:
ECKSTEIN, J;BERTSEKAS, DP
通讯作者: BERTSEKAS, DP
DOI: 10.1109/msp.2010.939739
发表时间: 2011-03
影响因子: 14.9
作者:
René Vidal
通讯作者: René Vidal
DOI: 10.1109/cvpr.2015.7298821
发表时间: 2015-06
期刊: 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
影响因子: --
作者:
Baohua Li;Ying Zhang;Zhouchen Lin;Huchuan Lu
通讯作者: Baohua Li;Ying Zhang;Zhouchen Lin;Huchuan Lu
DOI: 10.1002/cpa.20303
发表时间: 2010-01-01
影响因子: 3
作者:
Daubechies, Ingrid;Devore, Ronald;Guentuerk, C. Sinan
通讯作者: Guentuerk, C. Sinan