Geometric algebra for pose estimation and surface morphing in human motion estimation

  • Authors:
  • Bodo Rosenhahn;Reinhard Klette

  • Affiliations:
  • Computer Science Department, University of Auckland (CITR), Auckland, New Zealand;Computer Science Department, University of Auckland (CITR), Auckland, New Zealand

  • Venue:
  • IWCIA'04 Proceedings of the 10th international conference on Combinatorial Image Analysis
  • Year:
  • 2004

Quantified Score

Hi-index 0.00

Visualization

Abstract

We exploit properties of geometric algebras (GAs) to model the 2D-3D pose estimation problem for free-form surfaces which are coupled with kinematic chains. We further describe local and global surface morphing approaches with GA and combine them with the 2D-3D pose estimation problem. As an application of the presented approach, human motion estimation is considered. The estimated joint angles are used to deform surface patches to gain more realistic human models and therefore more accurate pose estimation results.