Spectral clustering of linear subspaces for motion segmentation - Archive ouverte HAL Access content directly
Conference Papers Year : 2009

Spectral clustering of linear subspaces for motion segmentation

Abstract

This paper studies automatic segmentation of multiple motions from tracked feature points through spectral embedding and clustering of linear subspaces. We show that the dimension of the ambient space is crucial for separability, and that low dimensions chosen in prior work are not optimal. We suggest lower and upper bounds together with a data-driven procedure for choosing the optimal ambient dimension. Application of our approach to the Hopkins155 video benchmark database uniformly outperforms a range of state-of-the-art methods both in terms of segmentation accuracy and computational speed.
Fichier principal
Vignette du fichier
LauerICCV09.pdf (752.2 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

hal-00396782 , version 1 (18-06-2009)

Identifiers

  • HAL Id : hal-00396782 , version 1

Cite

Fabien Lauer, Christoph Schnörr. Spectral clustering of linear subspaces for motion segmentation. IEEE Int. Conf. on Computer Vision (ICCV'09), Sep 2009, Kyoto, Japan. ⟨hal-00396782⟩
228 View
870 Download

Share

Gmail Facebook X LinkedIn More