August 2009
Volume 9, Issue 8
Vision Sciences Society Annual Meeting Abstract  |   August 2009
Extracting motion contours with simultaneous local and global processing mechanisms
Author Affiliations
  • Andrew Meso
    Psychology, Royal Holloway University of London
  • Andrew Shaw
    Psychology, Royal Holloway University of London
  • Szonya Durant
    Psychology, Royal Holloway University of London
  • Johannes Zanker
    Psychology, Royal Holloway University of London
Journal of Vision August 2009, Vol.9, 652. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Andrew Meso, Andrew Shaw, Szonya Durant, Johannes Zanker; Extracting motion contours with simultaneous local and global processing mechanisms. Journal of Vision 2009;9(8):652.

      Download citation file:

      © ARVO (1962-2015); The Authors (2016-present)

  • Supplements

Dynamic visual scenes present an observer with motion contours that indicate the location of the boundaries of moving objects. Because motion contours are useful for figure-ground discrimination and for the analysis of depth structure, their fast and accurate identification is ecologically significant. Visibility depends on the difference between motion vectors on either side of the contour, each of which has a speed and a direction. The relative contribution of speed and direction in determining the visibility of a motion-defined contour can provide important clues about the neural mechanisms underlying the perception of motion discontinuities. Here, we explore the computational requirements of detecting motion contours for stimuli that we previously investigated with psychophysical methods in a study which found that speed and direction are detected independently by human observers and combined such as to optimise perceptual performance (Durant and Zanker 2008, Vision Research 48, 1053–1060). We simulate the detection of motion contours by computing local motion signals using correlation detectors and deriving global motion patterns from the local signal distributions. From histograms of local motion signals, clusters corresponding to different regions of uniform motion are identified. The clusters are used to group local motion signals in order to segment the images and identify contours. This process is based on a hierarchical structure with forward and backward connectivity computing an initial local detection, then computing global motion to support a subsequent segmentation. The reliability of separating clusters attributable to the different stimulus regions is used as an indicator of the visibility of the contours. In computer simulations, we find that differences in direction are more reliably detected than differences in speed. We discuss this result and the general structure of the model in relation to the previous psychophysical findings.

Meso, A. Shaw, A. Durant, S. Zanker, J. (2009). Extracting motion contours with simultaneous local and global processing mechanisms [Abstract]. Journal of Vision, 9(8):652, 652a,, doi:10.1167/9.8.652. [CrossRef]
 With support from EPSRC (EP/0500 2329 and EP/C015061/1).

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.