Design and Use of Linear Models for Image Motion Analysis

  • Authors:

  • David J. Fleet;Michael J. Black;Yaser Yacoob;Allan D. Jepson

  • Affiliations:

  • Department of Computing and Information Science, Queen's University, Kingston, Ontario, Canada, K7L 3N6&semi/ Xerox Palo Alto Research Center, 3333 Coyote Hill Road, Palo Alto, CA 94304, USA. flee ...;Xerox Palo Alto Research Center, 3333 Coyote Hill Road, Palo Alto, CA 94304, USA. black@parc.xerox.com;Computer Vision Laboratory, University of Maryland, College Park, MD 20742, USA. yaser@cs.umd.edu;Department of Computer Science, University of Toronto, Toronto, Ontario, Canada, M5S 1A4. jepson@vis.toronto.edu

  • Venue:
  • International Journal of Computer Vision
  • Year:
  • 2000

Quantified Score

Hi-index 0.00

Visualization

Abstract

Linear parameterized models of optical flow, particularlyaffine models, have become widespread in image motion analysis.The linear model coefficients are straightforward to estimate, andthey provide reliable estimates of the optical flow of smoothsurfaces. Here we explore the use of parameterized motion models thatrepresent much more varied and complex motions. Our goals arethreefold: to construct linear bases for complex motion phenomena; toestimate the coefficients of these linear models; and to recognize orclassify image motions from the estimated coefficients. We considertwo broad classes of motions: i) generic “motion features”such as motion discontinuities and moving bars; and ii)non-rigid, object-specific, motions such as the motion of humanmouths. For motion features we construct a basis of steerableflow fields that approximate the motion features. Forobject-specific motions we construct basis flow fields from examplemotions using principal component analysis. In both cases, the modelcoefficients can be estimated directly from spatiotemporal imagederivatives with a robust, multi-resolution scheme. Finally, we showhow these model coefficients can be use to detect and recognizespecific motions such as occlusion boundaries and facialexpressions.