Early Scene Analysis: Rapid Procession of Contours, Surfaces and Objects in Human Vision

Abstract

Peter Murphy and I have created an image analysis program which labels image discontinuities as either occlusions, shadow, or pigment boundaries. The algorithm relies on regularities in the brightness flow surrounding each different contour type--each of contour has a characteristic form of isophotes in its neighborhood. For example, the isophotes of a surface lie parallel to an occluding boundary on the side of the nearer surface and have no particular relation to the boundary on the occluded surface. The isophotes of a surface are collinear across any pigment change, independently of the orientation. For some ambient lights though not all, the same relation holds for the isophotes across a cast shadow boundary. Highlights and attached shadows have additional characteristics forms. Due to Peter's untimely loss to a lucrative computer science job, this project is on hold.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Sep 01, 2000
Accession Number
ADA383742

Entities

People

  • Patrick Cavanagh

Organizations

  • Harvard University

Tags

Communities of Interest

  • Advanced Electronics
  • Human Systems

DTIC Thesaurus Topics

  • Algorithms
  • Biological Sciences
  • Boundaries
  • Brightness
  • Cognitive Science
  • Computer Science
  • Computer Vision
  • Computers
  • Identification
  • Neurosciences
  • Object Recognition
  • Orientation (Direction)
  • Perception
  • Psychology
  • Recognition
  • Three Dimensional
  • Two Dimensional

Readers

  • Electromagnetic Wave Scattering and Antenna Radiation Engineering
  • Finite Element Method (FEM) for solving Partial Differential Equations (PDEs)
  • Vision Science/Vision Psychology/Cognitive Neuroscience.