Robust PCA Based on Incoherence With Geometrical Interpretation

Jieun Lee, Yoonsik Choe

Research output: Contribution to journalArticle

2 Citations (Scopus)

Abstract

Robust principal component analysis, which extracts low-dimensional data from high-dimensional data, can also be regarded as a source separation problem of the sparse error matrix and the low-rank matrix. Until recently, various methods have attempted to precisely predict the discrete rank function by assigning a weight to the nuclear norm. However, if the weights are not in ascending order, the algorithms will diverge and exhibit high computational complexity. Moreover, from the viewpoint of source separation, these methods overlook the fact that two components must be sufficiently different for accurate demixing. In this paper, we employ the incoherence term with convex shape, which considers that components must appear different from one another for boosting separability. Since it is intractable to directly exploit mutual incoherence defined in linear algebra, we guarantee the incoherence by indirectly making the sparse matrix lack the low-rank property by using the duality norm principle. This approach can also be associated with the null space. To analyze the results of the proposed algorithm geometrically, we measure the geodesic distance between the tangent spaces of the manifolds of two separate components. As this distance increases, the degree of dissimilarity of the two components is adequately assured; thus, separation succeeds. Furthermore, this paper is the first to provide insights into the relationship between source separation conditions and the derivatives of the nuclear norm and L1 norm. Experiments are conducted on still image separation and background subtraction to confirm the superiority of the proposed methods both qualitatively and quantitatively.

Original languageEnglish
Article number8249529
Pages (from-to)1939-1950
Number of pages12
JournalIEEE Transactions on Image Processing
Volume27
Issue number4
DOIs
Publication statusPublished - 2018 Apr

Fingerprint

Source separation
Linear algebra
Principal component analysis
Computational complexity
Derivatives
Experiments

All Science Journal Classification (ASJC) codes

  • Software
  • Computer Graphics and Computer-Aided Design

Cite this

@article{f8bd8b864f9f42c781720cac55c28489,
title = "Robust PCA Based on Incoherence With Geometrical Interpretation",
abstract = "Robust principal component analysis, which extracts low-dimensional data from high-dimensional data, can also be regarded as a source separation problem of the sparse error matrix and the low-rank matrix. Until recently, various methods have attempted to precisely predict the discrete rank function by assigning a weight to the nuclear norm. However, if the weights are not in ascending order, the algorithms will diverge and exhibit high computational complexity. Moreover, from the viewpoint of source separation, these methods overlook the fact that two components must be sufficiently different for accurate demixing. In this paper, we employ the incoherence term with convex shape, which considers that components must appear different from one another for boosting separability. Since it is intractable to directly exploit mutual incoherence defined in linear algebra, we guarantee the incoherence by indirectly making the sparse matrix lack the low-rank property by using the duality norm principle. This approach can also be associated with the null space. To analyze the results of the proposed algorithm geometrically, we measure the geodesic distance between the tangent spaces of the manifolds of two separate components. As this distance increases, the degree of dissimilarity of the two components is adequately assured; thus, separation succeeds. Furthermore, this paper is the first to provide insights into the relationship between source separation conditions and the derivatives of the nuclear norm and L1 norm. Experiments are conducted on still image separation and background subtraction to confirm the superiority of the proposed methods both qualitatively and quantitatively.",
author = "Jieun Lee and Yoonsik Choe",
year = "2018",
month = "4",
doi = "10.1109/TIP.2018.2790481",
language = "English",
volume = "27",
pages = "1939--1950",
journal = "IEEE Transactions on Image Processing",
issn = "1057-7149",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "4",

}

Robust PCA Based on Incoherence With Geometrical Interpretation. / Lee, Jieun; Choe, Yoonsik.

In: IEEE Transactions on Image Processing, Vol. 27, No. 4, 8249529, 04.2018, p. 1939-1950.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Robust PCA Based on Incoherence With Geometrical Interpretation

AU - Lee, Jieun

AU - Choe, Yoonsik

PY - 2018/4

Y1 - 2018/4

N2 - Robust principal component analysis, which extracts low-dimensional data from high-dimensional data, can also be regarded as a source separation problem of the sparse error matrix and the low-rank matrix. Until recently, various methods have attempted to precisely predict the discrete rank function by assigning a weight to the nuclear norm. However, if the weights are not in ascending order, the algorithms will diverge and exhibit high computational complexity. Moreover, from the viewpoint of source separation, these methods overlook the fact that two components must be sufficiently different for accurate demixing. In this paper, we employ the incoherence term with convex shape, which considers that components must appear different from one another for boosting separability. Since it is intractable to directly exploit mutual incoherence defined in linear algebra, we guarantee the incoherence by indirectly making the sparse matrix lack the low-rank property by using the duality norm principle. This approach can also be associated with the null space. To analyze the results of the proposed algorithm geometrically, we measure the geodesic distance between the tangent spaces of the manifolds of two separate components. As this distance increases, the degree of dissimilarity of the two components is adequately assured; thus, separation succeeds. Furthermore, this paper is the first to provide insights into the relationship between source separation conditions and the derivatives of the nuclear norm and L1 norm. Experiments are conducted on still image separation and background subtraction to confirm the superiority of the proposed methods both qualitatively and quantitatively.

AB - Robust principal component analysis, which extracts low-dimensional data from high-dimensional data, can also be regarded as a source separation problem of the sparse error matrix and the low-rank matrix. Until recently, various methods have attempted to precisely predict the discrete rank function by assigning a weight to the nuclear norm. However, if the weights are not in ascending order, the algorithms will diverge and exhibit high computational complexity. Moreover, from the viewpoint of source separation, these methods overlook the fact that two components must be sufficiently different for accurate demixing. In this paper, we employ the incoherence term with convex shape, which considers that components must appear different from one another for boosting separability. Since it is intractable to directly exploit mutual incoherence defined in linear algebra, we guarantee the incoherence by indirectly making the sparse matrix lack the low-rank property by using the duality norm principle. This approach can also be associated with the null space. To analyze the results of the proposed algorithm geometrically, we measure the geodesic distance between the tangent spaces of the manifolds of two separate components. As this distance increases, the degree of dissimilarity of the two components is adequately assured; thus, separation succeeds. Furthermore, this paper is the first to provide insights into the relationship between source separation conditions and the derivatives of the nuclear norm and L1 norm. Experiments are conducted on still image separation and background subtraction to confirm the superiority of the proposed methods both qualitatively and quantitatively.

UR - http://www.scopus.com/inward/record.url?scp=85040560613&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85040560613&partnerID=8YFLogxK

U2 - 10.1109/TIP.2018.2790481

DO - 10.1109/TIP.2018.2790481

M3 - Article

AN - SCOPUS:85040560613

VL - 27

SP - 1939

EP - 1950

JO - IEEE Transactions on Image Processing

JF - IEEE Transactions on Image Processing

SN - 1057-7149

IS - 4

M1 - 8249529

ER -