Three-Dimensional Human Body Model Acquisition from Multiple Views

  • Authors:

  • Ioannis A. Kakadiaris;Dimitri Metaxas

  • Affiliations:

  • Department of Computer and Information Science, University of Pennsylvania, Philadelphia, PA 19104-6389. E-mail: ioannisk@grip.cis.upenn.edu;Department of Computer and Information Science, University of Pennsylvania, Philadelphia, PA 19104-6389. E-mail: dnm@central.cis.upenn.edu

  • Venue:
  • International Journal of Computer Vision
  • Year:
  • 1998

Quantified Score

Hi-index 0.00

Visualization

Abstract

We present a novel approach to the three-dimensional humanbody model acquisition from three mutually orthogonal views. Ourtechnique is based on the spatiotemporal analysis of the deformingapparent contour of a human moving according to a protocol ofmovements. For generality and robustness our technique does not use aprior model of the human body and a prior body part segmentation isnot assumed. Therefore, our technique applies to humans of anyanthropometric dimension. To parameterize and segment over time adeforming apparent contour, we introduce a new shape representationtechnique based on primitive composition. The composeddeformable model allows us to represent large local deformations andtheir evolution in a compact and intuitive way. In addition, thisrepresentation allows us to hypothesize an underlying part structureand test this hypothesis against the relative motion (due to forcesexerted from the image data) of the defining primitives of thecomposed model. Furthermore, we develop a Human Body PartDecomposition Algorithm (HBPDA) that recovers all the body parts ofa subject by monitoring the changes over time to the shape of thedeforming silhouette. In addition, we modularize the process ofsimultaneous two-dimensional part determination and shape estimationby employing the Supervisory Control Theory of Discrete EventSystems. Finally, we present a novel algorithm which selectivelyintegrates the (segmented by the HBPDA) apparent contours from threemutually orthogonal viewpoints to obtain a three-dimensional model ofthe subject‘s body parts. The effectiveness of the approach isdemonstrated through a series of experiments where a subject performsa set of movements according to a protocol that reveals the structureof the human body.