Fast-robust PCA

Markus Storer, Peter Roth, Martin Urschler, Horst Bischof

Research output: Chapter in Book/Report/Conference proceedingChapterResearch

Abstract

Principal Component Analysis (PCA) is a powerful and widely used tool in Computer Vision and is applied, e.g., for dimensionality reduction. But as a drawback, it is not robust to outliers. Hence, if the input data is corrupted, an arbitrarily wrong representation is obtained. To overcome this problem, various methods have been proposed to robustly estimate the PCA coefficients, but these methods are computationally too expensive for practical applications. Thus, in this paper we propose a novel fast and robust PCA (FR-PCA), which drastically reduces the computational effort. Moreover, more accurate representations are obtained. In particular, we propose a two-stage outlier detection procedure, where in the first stage outliers are detected by analyzing a large number of smaller subspaces. In the second stage, remaining outliers are detected by a robust least-square fitting. To show these benefits, in the experiments we evaluate the FR-PCA method for the task of robust image reconstruction on the publicly available ALOI database. The results clearly show that our approach outperforms existing methods in terms of accuracy and speed when processing corrupted data.
Original languageEnglish
Title of host publicationImage Analysis
Subtitle of host publication16th Scandinavian Conference, SCIA 2009, Oslo, Norway, June 15-18, 2009. Proceedings
EditorsArnt-Børre Salberg, Jon Y. Hardeberg, Robert Jensen
Place of PublicationBerlin Heidelberg
PublisherSpringer
Pages430-439
Volume5575
ISBN (Electronic)978-3-642-02230-2
ISBN (Print)978-3-642-02229-6
DOIs
Publication statusPublished - 2009

Fingerprint

Principal component analysis
Image reconstruction
Computer vision
Experiments

Fields of Expertise

  • Information, Communication & Computing

Cite this

Storer, M., Roth, P., Urschler, M., & Bischof, H. (2009). Fast-robust PCA. In A-B. Salberg, J. Y. Hardeberg, & R. Jensen (Eds.), Image Analysis: 16th Scandinavian Conference, SCIA 2009, Oslo, Norway, June 15-18, 2009. Proceedings (Vol. 5575, pp. 430-439). Berlin Heidelberg: Springer. https://doi.org/10.1007/978-3-642-02230-2_44

Fast-robust PCA. / Storer, Markus; Roth, Peter; Urschler, Martin; Bischof, Horst.

Image Analysis: 16th Scandinavian Conference, SCIA 2009, Oslo, Norway, June 15-18, 2009. Proceedings. ed. / Arnt-Børre Salberg; Jon Y. Hardeberg; Robert Jensen. Vol. 5575 Berlin Heidelberg : Springer, 2009. p. 430-439.

Research output: Chapter in Book/Report/Conference proceedingChapterResearch

Storer, M, Roth, P, Urschler, M & Bischof, H 2009, Fast-robust PCA. in A-B Salberg, JY Hardeberg & R Jensen (eds), Image Analysis: 16th Scandinavian Conference, SCIA 2009, Oslo, Norway, June 15-18, 2009. Proceedings. vol. 5575, Springer, Berlin Heidelberg, pp. 430-439. https://doi.org/10.1007/978-3-642-02230-2_44
Storer M, Roth P, Urschler M, Bischof H. Fast-robust PCA. In Salberg A-B, Hardeberg JY, Jensen R, editors, Image Analysis: 16th Scandinavian Conference, SCIA 2009, Oslo, Norway, June 15-18, 2009. Proceedings. Vol. 5575. Berlin Heidelberg: Springer. 2009. p. 430-439 https://doi.org/10.1007/978-3-642-02230-2_44
Storer, Markus ; Roth, Peter ; Urschler, Martin ; Bischof, Horst. / Fast-robust PCA. Image Analysis: 16th Scandinavian Conference, SCIA 2009, Oslo, Norway, June 15-18, 2009. Proceedings. editor / Arnt-Børre Salberg ; Jon Y. Hardeberg ; Robert Jensen. Vol. 5575 Berlin Heidelberg : Springer, 2009. pp. 430-439
@inbook{5118331f77fa4d58992340f98229a413,
title = "Fast-robust PCA",
abstract = "Principal Component Analysis (PCA) is a powerful and widely used tool in Computer Vision and is applied, e.g., for dimensionality reduction. But as a drawback, it is not robust to outliers. Hence, if the input data is corrupted, an arbitrarily wrong representation is obtained. To overcome this problem, various methods have been proposed to robustly estimate the PCA coefficients, but these methods are computationally too expensive for practical applications. Thus, in this paper we propose a novel fast and robust PCA (FR-PCA), which drastically reduces the computational effort. Moreover, more accurate representations are obtained. In particular, we propose a two-stage outlier detection procedure, where in the first stage outliers are detected by analyzing a large number of smaller subspaces. In the second stage, remaining outliers are detected by a robust least-square fitting. To show these benefits, in the experiments we evaluate the FR-PCA method for the task of robust image reconstruction on the publicly available ALOI database. The results clearly show that our approach outperforms existing methods in terms of accuracy and speed when processing corrupted data.",
author = "Markus Storer and Peter Roth and Martin Urschler and Horst Bischof",
note = "Volume 5575 of LNCS",
year = "2009",
doi = "10.1007/978-3-642-02230-2_44",
language = "English",
isbn = "978-3-642-02229-6",
volume = "5575",
pages = "430--439",
editor = "Arnt-B{\o}rre Salberg and Hardeberg, {Jon Y.} and Robert Jensen",
booktitle = "Image Analysis",
publisher = "Springer",

}

TY - CHAP

T1 - Fast-robust PCA

AU - Storer, Markus

AU - Roth, Peter

AU - Urschler, Martin

AU - Bischof, Horst

N1 - Volume 5575 of LNCS

PY - 2009

Y1 - 2009

N2 - Principal Component Analysis (PCA) is a powerful and widely used tool in Computer Vision and is applied, e.g., for dimensionality reduction. But as a drawback, it is not robust to outliers. Hence, if the input data is corrupted, an arbitrarily wrong representation is obtained. To overcome this problem, various methods have been proposed to robustly estimate the PCA coefficients, but these methods are computationally too expensive for practical applications. Thus, in this paper we propose a novel fast and robust PCA (FR-PCA), which drastically reduces the computational effort. Moreover, more accurate representations are obtained. In particular, we propose a two-stage outlier detection procedure, where in the first stage outliers are detected by analyzing a large number of smaller subspaces. In the second stage, remaining outliers are detected by a robust least-square fitting. To show these benefits, in the experiments we evaluate the FR-PCA method for the task of robust image reconstruction on the publicly available ALOI database. The results clearly show that our approach outperforms existing methods in terms of accuracy and speed when processing corrupted data.

AB - Principal Component Analysis (PCA) is a powerful and widely used tool in Computer Vision and is applied, e.g., for dimensionality reduction. But as a drawback, it is not robust to outliers. Hence, if the input data is corrupted, an arbitrarily wrong representation is obtained. To overcome this problem, various methods have been proposed to robustly estimate the PCA coefficients, but these methods are computationally too expensive for practical applications. Thus, in this paper we propose a novel fast and robust PCA (FR-PCA), which drastically reduces the computational effort. Moreover, more accurate representations are obtained. In particular, we propose a two-stage outlier detection procedure, where in the first stage outliers are detected by analyzing a large number of smaller subspaces. In the second stage, remaining outliers are detected by a robust least-square fitting. To show these benefits, in the experiments we evaluate the FR-PCA method for the task of robust image reconstruction on the publicly available ALOI database. The results clearly show that our approach outperforms existing methods in terms of accuracy and speed when processing corrupted data.

U2 - 10.1007/978-3-642-02230-2_44

DO - 10.1007/978-3-642-02230-2_44

M3 - Chapter

SN - 978-3-642-02229-6

VL - 5575

SP - 430

EP - 439

BT - Image Analysis

A2 - Salberg, Arnt-Børre

A2 - Hardeberg, Jon Y.

A2 - Jensen, Robert

PB - Springer

CY - Berlin Heidelberg

ER -