Share Email Print

Proceedings Paper

Dynamical systems approach to low-level integration of motion and binocular stereopsis
Author(s): Edward J. Altman
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

The representation of trajectories for moving edges by a nonlinear dynamical system is presented as a technique for the integration of motion and binocular disparity. The integration is accomplished through the monolithic use of dynamical systems as the computational mechanism. In this approach, the motion of a simple feature through a small region of the image is modeled by a dynamical system with three variables. Visual processing by the model dynamics occurs within a retinotopic array of identical dynamical systems with nearest neighbor coupling. A form of segmentation, called feature linking, is accomplished by the convergence of neighboring array elements to the same stable mode of coherent oscillations involving only those dynamical systems that respond to the co-occurrence of the same feature and motion. Each model dynamic for a moving feature is driven by concurrent inputs from a pair of Gabor filters in phase quadrature. The stable modes of the model dynamics consist of limit cycles with nearly linear phase. The co-occurrence of a particular feature and velocity of motion in the image sequence optimally selects a particular limit cycle. The integration of motion and binocular stereopsis is achieved through the use of the linear phase property of the limit cycles to recover the positional disparity between similar features moving through corresponding regions in the retinotopic arrays for the left and right images. Bifurcation theory is applied to rigorously describe the convergence properties of the dynamical systems. The focus of this paper concerns the mapping of visual motion onto analog dynamical systems as an initial step toward a theory of low-level integration of motion and binocular stereopsis.

Paper Details

Date Published: 1 February 1992
PDF: 11 pages
Proc. SPIE 1610, Curves and Surfaces in Computer Vision and Graphics II, (1 February 1992); doi: 10.1117/12.135160
Show Author Affiliations
Edward J. Altman, Univ. of Illinois/Urbana-Champaign (United States)

Published in SPIE Proceedings Vol. 1610:
Curves and Surfaces in Computer Vision and Graphics II
Martine J. Silbermann; Hemant D. Tagare, Editor(s)

© SPIE. Terms of Use
Back to Top