Connectionnist Algorithm for a 3D Dense Image Building from Stereoscopy

  • Marc-Noël Fauvel
  • Pascal Aubry
Conference paper


In this paper, we present a neuron-like network able to build 3D dense maps from stereoscopic image pair. The process of stereopsis is encoded by an energy function, which controls the evolving of the network. This one has the same structure than original images, and evolutes on a simple gradient steep. Thus, the system is fully parallel and could be hardware implemented for a real time use.


Energy Function Outdoor Scene Correlation Window Coherence Term Teleoperated Robot 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. [1]
    N. Ayache, B. Faverjon: Un algorithme rapide de stéréoscopie passive utilisant la prédiction et vérification récursive d’hypothèse. 5è congrès de reconnaissance des formes et intelligence artificielle: Grenoble 1985.Google Scholar
  2. [2]
    H.H. Bake: Depth from edge & intensity based stereo. Technical repport AIM-347, Stanford University: California, September 1982.Google Scholar
  3. [3]
    S.D. Cochran: Surface Description from binocular stereo. DARPA F33615-87-C-1436: Los Angeles, California, november 1990.Google Scholar
  4. [4]
    P. Fua: A parallel stereo algorithm that produces dense depth maps and preserves image features. INRIA rapport de recherche 1369: janvier 1991.Google Scholar
  5. [5]
    D. Marr, T. Poggio:, Cooperative computation of stereo disparity. Science 194: 283–287, 1976.CrossRefGoogle Scholar
  6. [6]
    J.J. Hopfield: Neural networks and physical systems with emergent collective computational abilities. Proceedings of the national academy of sciences 79: 2554–2558, 1982.MathSciNetCrossRefGoogle Scholar
  7. [7]
    J.J. Hopfield: Neurons with graded response have collective computational properties like those of 2-state. Proceedings of the national academy of sciences 81: 3088–3092, 1984.CrossRefGoogle Scholar

Copyright information

© Springer-Verlag/Wien 1995

Authors and Affiliations

  • Marc-Noël Fauvel
    • 1
  • Pascal Aubry
    • 1

Personalised recommendations