Image cues fusion for contour tracking based on particle filter - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Conference Papers Year : 2004

Image cues fusion for contour tracking based on particle filter

Abstract

Particle filter is a powerful algorithm to deal with non-linear and non-Gaussian tracking problems. However the algorithm relying only upon one image cue often fails in challenging scenarios. To overcome this, the paper first presents a color likelihood to capture color distribution of the object based on Bhattacharry coefficient, and a structure likelihood representing high level knowledge regarding the object. Together with the widely used edge likelihood, the paper further proposes a straightforward image cues fusion for object tracking in the framework of particle filter, under assumption that the visual measurement of each image cue is independent of each other. The experiments on real image sequences have shown that the method is effective, robust to illumination changes, pose variations and complex background.
Fichier principal
Vignette du fichier
2004_wkamdo_li.pdf (146.01 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

inria-00352027 , version 1 (12-01-2009)

Identifiers

  • HAL Id : inria-00352027 , version 1

Cite

P. Li, François Chaumette. Image cues fusion for contour tracking based on particle filter. Int. Workshop on Articulated Motion and Deformable Objects, AMDO'04, 2004, Palma de Mallorca, Spain. pp.99-107. ⟨inria-00352027⟩
165 View
143 Download

Share

Gmail Facebook X LinkedIn More