An M-estimator for robust centroid estimation on the manifold of covariance matrices: performance analysis and application to image classification

Abstract : Many signal and image processing applications, including texture analysis, radar detection or EEG signal classification, require the computation of a centroid from a set of covariance matrices. The most popular approach consists in considering the center of mass. While efficient, this estimator is not robust to outliers arising from the inherent variability of the data or from faulty measurements. To overcome this, some authors have proposed to use the median as a more robust estimator. Here, we propose an estimator which takes advantage of both efficiency and robustness by combining the concepts of Riemannian center of mass and median. Based on the theory of M-estimators, this robust centroid estimator is issued from the so-called Huber's function. We present a gradient descent algorithm to estimate it. In addition, an experiment on both simulated and real data is carried out to evaluate the influence of outliers on the estimation and classification performances.
Document type :
Conference papers
Complete list of metadatas

Cited literature [30 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-01379604
Contributor : Lionel Bombrun <>
Submitted on : Tuesday, October 11, 2016 - 5:30:21 PM
Last modification on : Monday, November 26, 2018 - 1:30:03 PM
Long-term archiving on : Saturday, February 4, 2017 - 7:27:00 PM

File

Ilea16_EUSIPCO.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01379604, version 1

Citation

Ioana Ilea, Hatem Hajri, Salem Said, Lionel Bombrun, Christian Germain, et al.. An M-estimator for robust centroid estimation on the manifold of covariance matrices: performance analysis and application to image classification. 24th European Signal Processing Conference (EUSIPCO'16), Aug 2016, Budapest, Hungary. pp.2196 - 2200. ⟨hal-01379604⟩

Share

Metrics

Record views

252

Files downloads

206