Computation of Stereo and Visual Motion: From Biophysics to Psychophysics

Abstract

During the second and third quarter of this stretchout year of funding, we have continued to explore a number of problems in motion analysis, including the parallel detection of motion using a correlation-based mechanism, motion correspondence, neural mechanisms for motion detection and measurement, and the recovery of 3-D structure and motion. We are also starting to focus more deeply on the integration of multiple visual cues. Described here is some work on the interaction between surface shape, albedo, and the illuminant direction. As we noted in our previous report, we are developing and testing some variations on a parallel network model recently proposed by Hutchinson, Koch, Luo and Mead for combining the computation of the smoothest velocity field with line processes (suggested by Geman and Geman) for handling motion discontinuities. Our modified network derives the initial motion measurements only at the locations of significant intensity changes, allows greater flexibility in the placement of the discontinuities and considers variations on the energy function being minimized to implement the smoothness constraint. The network is also designed in a way that more closely parallels physiological properties of motion-sensitive neurons in area MT of monkey visual cortex.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Sep 30, 1988
Accession Number
ADA201873

Entities

People

  • Ellen Hildreth
  • Heinrich Buelthoff
  • Norberto Grzywacz
  • Tomaso Poggio

Organizations

  • Massachusetts Institute of Technology

Tags

Communities of Interest

  • Air Platforms
  • Energy and Power Technologies
  • Materials and Manufacturing Processes

DTIC Thesaurus Topics

  • Algorithms
  • Computational Science
  • Computations
  • Computer Stereo Vision
  • Data Analysis
  • Detectors
  • Discontinuities
  • Geometry
  • Measurement
  • Models
  • Observers
  • Perception
  • Stratified Fluids
  • Three Dimensional
  • Trajectories
  • Two Dimensional
  • Visual Cortex

Readers

  • Approximation Theory.
  • Computer Vision.
  • Neural Network Machine Learning.

Technology Areas

  • Biotechnology