August 2012
Volume 12, Issue 9
Free
Vision Sciences Society Annual Meeting Abstract  |   August 2012
Emergent features in object detection
Author Affiliations
  • Matthew Inverso
    Department of Psychology, Rutgers University\nRutgers University Center for Cognitive Sciences
  • John Wilder
    Department of Psychology, Rutgers University\nRutgers University Center for Cognitive Sciences
  • Jacob Feldman
    Department of Psychology, Rutgers University\nRutgers University Center for Cognitive Sciences
  • Manish Singh
    Department of Psychology, Rutgers University\nRutgers University Center for Cognitive Sciences
Journal of Vision August 2012, Vol.12, 886. doi:10.1167/12.9.886
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Matthew Inverso, John Wilder, Jacob Feldman, Manish Singh; Emergent features in object detection. Journal of Vision 2012;12(9):886. doi: 10.1167/12.9.886.

      Download citation file:


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

      ×
  • Supplements
Abstract

Detection of simple isolated contours amid random background elements is a basic perceptual capacity, and has been extensively studied. But while simple contours are essential building blocks of object representations, objects and scenes also involve emergent or "Gestalt" configural features that involve nonlocal combinations of contours. Here we investigate how such emergent features contribute to object detection. We constructed target objects from pairs of simple contours in various spatial relationships, and manipulated both the structure of the individual contours as well as the angle between the two contours -- a simple example of an emergent relational feature. The targets were embedded in grayscale pixel noise fields and subjects were asked to detect them, in a both an accuracy paradigm (2IFC, Exp. 1) and a response latency paradigm (2AFC, Exp. 2). In previous studies in our lab (Wilder, Feldman & Singh, VSS 2011) we demonstrated a strong influence of the complexity of individual contours on detectability, where complexity is quantified via the description length (DL) of the contour, defined as the cumulative surprisal (-log p) of the sequence of turning angles that make it up. In the current studies, we again find a strong effect of the complexity of the individual contours. But here we also find an independent effect of the angle between the two contours, with enhanced performance at both parallel (0 deg) and perpendicular (90 deg) configurations, and depressed performance at intermediate angles. This effect can be incorporated directly into the complexity formulation simply by adopting a mixture prior over the inter-contour angle and adding its surprisal to the DL of the ensemble. These results contribute to a more complete understanding of the how local and global features are integrated into object representations.

Meeting abstract presented at VSS 2012

×
×

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.

×