Advertisement
Research Article

Object Segmentation from Motion Discontinuities and Temporal Occlusions–A Biologically Inspired Model

  • Cornelia Beck mail,

    cornelia.beck@uni-ulm.de

    Affiliation: Institute for Neural Information Processing, University of Ulm, Ulm, Germany

    X
  • Thilo Ognibeni,

    Affiliation: Institute for Neural Information Processing, University of Ulm, Ulm, Germany

    X
  • Heiko Neumann

    Affiliation: Institute for Neural Information Processing, University of Ulm, Ulm, Germany

    X
  • Published: November 27, 2008
  • DOI: 10.1371/journal.pone.0003807

Reader Comments (2)

Post a new comment on this article

Capturing the Flow – Ernest Greene, Academic Editor

Posted by egreene on 05 Jan 2009 at 17:08 GMT

These authors propose a method for staged-processing of motion information that serves to distinguish the boundaries of a moving object, and thus allow the object itself to be segregated. The model includes robust optic flow estimation along motion boundaries, detection of motion discontinuities with respect to those boundaries, and the analysis of occlusion differentials as one object passes in front of another. Representations of motion discontinuities and occlusion evidences mutually interact to improve and stabilize the results. The authors draw parallels among the several stages of processing, and what appear to be comparable functions of several brain regions. Their model achieved object segmentation in both artificial and natural scenes.