Convergence of iteratively re-weighted least squares to robust M-estimators

Khurrum Aftab, Richard Hartley

    Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

    46 Citations (Scopus)

    Abstract

    This paper presents a way of using the Iteratively Reweighted Least Squares (IRLS) method to minimize several robust cost functions such as the Huber function, the Cauchy function and others. It is known that IRLS (otherwise known as Weiszfeld) techniques are generally more robust to outliers than the corresponding least squares methods, but the full range of robust M-estimators that are amenable to IRLS has not been investigated. In this paper we address this question and show that IRLS methods can be used to minimize most common robust M-estimators. An exact condition is given and proved for decrease of the cost, from which convergence follows. In addition to the advantage of increased robustness, the proposed algorithm is far simpler than the standard L1 Weiszfeld algorithm. We show the applicability of the proposed algorithm to the rotation averaging, triangulation and point cloud alignment problems.

    Original languageEnglish
    Title of host publicationProceedings - 2015 IEEE Winter Conference on Applications of Computer Vision, WACV 2015
    PublisherInstitute of Electrical and Electronics Engineers Inc.
    Pages480-487
    Number of pages8
    ISBN (Electronic)9781479966820
    DOIs
    Publication statusPublished - 19 Feb 2015
    Event2015 15th IEEE Winter Conference on Applications of Computer Vision, WACV 2015 - Waikoloa, United States
    Duration: 5 Jan 20159 Jan 2015

    Publication series

    NameProceedings - 2015 IEEE Winter Conference on Applications of Computer Vision, WACV 2015

    Conference

    Conference2015 15th IEEE Winter Conference on Applications of Computer Vision, WACV 2015
    Country/TerritoryUnited States
    CityWaikoloa
    Period5/01/159/01/15

    Fingerprint

    Dive into the research topics of 'Convergence of iteratively re-weighted least squares to robust M-estimators'. Together they form a unique fingerprint.

    Cite this