An Integrated Neural and Algorithmic System for Optical Flow Computation

  • A. Criminisi
  • G. A. M. Gioiello
  • D. Molinelli
  • F. Sorbello
Conference paper
Part of the Perspectives in Neural Computing book series (PERSPECT.NEURAL)


Motion detection plays a central role in several visual environments: knowledge of object velocities and trajectories is fundamental in scene interpretation and segmentation. This task appears a simple problem, but detecting moving objects is very difficult, in fact this is a problem that cannot be considered completely solved today [1] [2] [3].

In this paper we present a novel method that uses two different approaches: a “neural” one and an algorithmic one. In fact, a Multilayer Perceptron is used in the first stage, in order to detect some motion areas in the scene [5] [6]; a matching algorithm is then used to obtain a sparse optical flow and to compute the epipolar geometry of the moving camera [7] [8]; and, finally, a refinement algorithm is used to produce a denser optical flow field. Thus this method can extract features automatically from moving objects in a scene discarding stationary ones. This approach seems to be very useful for tracking and motion segmentation.

This work was developed in the context of JACOB project, to achieve the automatic retrieval of images based on motion [9].


Optical Flow Fundamental Matrix Correlation Score Epipolar Line Epipolar Geometry 
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]
    S. Negahdaripour, S. Lee, Motion recovery from image sequences using only first order optical flow information, International Journal of computer vision, 9:3, 163–184 (1992)CrossRefGoogle Scholar
  2. [2]
    A. Singh, P. Allen, Image-Flow Computation An estimation-theoretic framework and an unified perspective, CVGIP-IU, vol. 56, no. 2, September 1992Google Scholar
  3. [3]
    J. Weber, J. Malik, Robust computation of optical flow in a multi-scale differential framework, IEEE Fourth International Conference on Computer Vision 2/93Google Scholar
  4. [4]
    M.J.D. Powel, Restart Procedures for the Conjugate Gradient Method, Mathematical Programming, Vol. 12, pp. 241–254Google Scholar
  5. [5]
    A. Abruzzo, G.A.M. Gioiello, M. La Cascia, F. Sorbello, Motion Detection From Image Sequences Using A New Fully Digital VLSI Neural Architecture, EUFIT95 - August 26–31, 1995, Aachen, GermanyGoogle Scholar
  6. [6]
    A. Abruzzo, G.A.M. Gioiello, M. La Cascia, F. Sorbello, A New Fully Digital Neural Architecture For Motion Detection From Image Sequences, EANN95 - International Conference on Engineering Applications of Neural Networks, August 21–23, 1995, Helsinki, FinlandGoogle Scholar
  7. [7]
    Q.T. Luong, R. Deriche, O.D. Faugeras, T. Papadopoulo, On Determining the Fundamental Matrix, Technical Report 1894, INULA, Sophia-Antipolis, France, 1993Google Scholar
  8. [8]
    Z. Zhang, R. Deriche, O. Faugeras, Q.-T. Luong, A Robust Technique for Matching Two Uncalibrated Images Through the Recovery of the Unknown Epipolar Geometry, Artificial Intelligence Journal, Vol.78, pages 87–119, October 1995. Also Research Report No.2273, INRIA Sophia-AntipolisCrossRefGoogle Scholar
  9. [9]
    M. La Cascia, E. Ardizzone, JACOB: Just a content-based query system for video databases, IEEE International Conference On ACOUSTICS, SPEECH AND SIGNAL PROCESSING, May 7–10, 1996, Atlanta, Georgia (USA)Google Scholar

Copyright information

© Springer-Verlag London Limited 1997

Authors and Affiliations

  • A. Criminisi
    • 1
  • G. A. M. Gioiello
    • 2
  • D. Molinelli
    • 1
  • F. Sorbello
    • 1
  1. 1.DIE - Dipartimento di Ingegneria ElettricaUniversità di PalermoPalermoItaly
  2. 2.DIE and CRESCentro per la Ricerca Elettronica in SiciliaMonrealeItaly

Personalised recommendations