Advertisement

Directions of motion fields are hardly ever ambiguous

  • Tomas Brodsky
  • Cornelia Fermüller
  • Yiannis Aloimonos
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 1065)

Abstract

Recent literature [7, 10, 11, 9, 13, 17] provides a number of results regarding uniqueness aspects of motion fields and exact image displacements due to 3-D rigid motion. Here instead of the full motion field we consider only the direction of the motion field due to a rigid motion and ask what can we say about the three-dimensional motion information contained in it. This paper provides a geometric analysis of this question based solely on the fact that the depth of the surfaces in view is positive (i.e. that the surface in view is in front of the camera). With this analysis we thus offer a theoretical foundation for image constraints employing only the sign of flow in various directions and provide a solid basis for their utilization in addressing 3D dynamic vision problems.

For two different rigid motions (with instantaneous translational and rotational velocities (t1, Ω1) and (t2, Ω2)) to yield the same direction of the flow, the surfaces in view must satisfy certain inequality and equality constraints, called critical surface constraints. A complete description of image areas where the constraints cannot be satisfied is derived and it is shown that if the imaging surface is a whole sphere, any two motions with different translation and rotation axes can be distinguished using only the direction of the flow. In the case where the imaging surface is a hemisphere or a plane, it is shown that two motions could give rise to the same direction of the flow if (t1×t2)·(Ω1×Ω2)=0 and several additional constraints are satisfied. For this to occur, the surfaces in view must satisfy all the critical surface constraints; thus at some points only a single depth value is allowed. Similar results are obtained for the case of multiple motions. Consequently, directions of motion fields are hardly ever ambiguous.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    N. Ancona and T. Poggio. Optical flow from 1-D correlation: Application to a simple time-to-crash detector. International Journal of Computer Vision: Special Issue on Qualitative Vision, Y. Aloimonos (Ed.), 14:131–146, 1995.Google Scholar
  2. 2.
    T. Brodsky, C. Fermüller and Y. Aloimonos. Directions of motion fields are hardly ever ambiguous. Technical Report CAR-TR-780, Center for Automation Research, University of Maryland, 1995.Google Scholar
  3. 3.
    C. Fermüller. Passive navigation as a pattern recognition problem. International Journal of Computer Vision: Special Issue on Qualitative Vision, Y. Aloimonos (Ed.), 14:147–158, 1995.Google Scholar
  4. 4.
    C. Fermüller and Y. Aloimonos. Direct Perception of Three-Dimensional Motion from Patterns of Visual Motion. Science, 270:1973–1976, 1995.Google Scholar
  5. 5.
    C. Fermüller and Y. Aloimonos. On the geometry of visual correspondence. International Journal of Computer Vision, 1995. To appear.Google Scholar
  6. 6.
    C. Fermüller and Y. Aloimonos. Qualitative egomotion. International Journal of Computer Vision, 15:7–29, 1995.Google Scholar
  7. 7.
    B. Horn. Motion fields are hardly ever ambiguous. International Journal of Computer Vision, 1:259–274, 1987.Google Scholar
  8. 8.
    B. Horn. Relative orientation. International Journal of Computer Vision, 4:59–78, 1990.Google Scholar
  9. 9.
    H. Longuet-Higgins. A computer algorithm for reconstruction of a scene from two projections. Nature, 293:133–135, 1981.Google Scholar
  10. 10.
    S. Maybank. Theory of Reconstruction from Image Motion. Springer, Berlin, 1993.Google Scholar
  11. 11.
    S. Negahdaripour. Critical surface pairs and triplets. International Journal of Computer Vision, 3:293–312, 1989.Google Scholar
  12. 12.
    S. Negahdaripour and B. Horn. Direct passive navigation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 9:163–176, 1987.Google Scholar
  13. 13.
    M. Spetsakis and J. Aloimonos. Structure from motion using line correspondences. International Journal of Computer Vision, 1:171–183, 1990.Google Scholar
  14. 14.
    M. Spetsakis and J. Aloimonos. A unified theory of structure from motion. In Proc. ARPA Image Understanding Workshop, Pittsburgh, PA, pages 271–283, 1990.Google Scholar
  15. 15.
    M. Spetsakis and J. Aloimonos. Optimal Visual Motion Estimation. IEEE Transactions on PAMI, 14:959–964, 1992.Google Scholar
  16. 16.
    M. Tistarelli and G. Sandini. Dynamic aspects in active vision. CVGIP: Image Understanding: Special Issue on Purposive, Qualitative, Active Vision, Y. Aloimonos (Ed.), 56:108–129, 1992.Google Scholar
  17. 17.
    R. Tsai and T. Huang. Uniqueness and estimation of three-dimensional motion parameters of rigid objects with curved surfaces. IEEE Transactions on Pattern Analysis and Machine Intelligence, 6:13–27, 1984.Google Scholar
  18. 18.
    T. Vieville and O. D. Faugeras. Robust and fast computation of edge characteristics in image sequences. International Journal of Computer Vision, 13:153–179, 1994.Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 1996

Authors and Affiliations

  • Tomas Brodsky
    • 1
  • Cornelia Fermüller
    • 1
  • Yiannis Aloimonos
    • 1
  1. 1.Computer Vision Laboratory, Center for Automation Research, Dept. of Computer Science and Institute of Advanced Computer StudiesUniversity of MarylandCollege Park

Personalised recommendations