August 2016
Volume 16, Issue 12
Open Access
Vision Sciences Society Annual Meeting Abstract  |   September 2016
Using the texture-centroid method to analyze the mechanisms sensitive to higher-order image statistics
Author Affiliations
  • Kier Groulx
    Department of Cognitive Sciences, University of California, Irvine
  • Charles Chubb
    Department of Cognitive Sciences, University of California, Irvine
  • Jonathan Victor
    Brain and Mind Research Institute, Weill Cornell Medical College
  • Mary Conte
    Brain and Mind Research Institute, Weill Cornell Medical College
Journal of Vision September 2016, Vol.16, 1033. doi:https://doi.org/10.1167/16.12.1033
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Kier Groulx, Charles Chubb, Jonathan Victor, Mary Conte; Using the texture-centroid method to analyze the mechanisms sensitive to higher-order image statistics. Journal of Vision 2016;16(12):1033. https://doi.org/10.1167/16.12.1033.

      Download citation file:


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

      ×
  • Supplements
Abstract

Rationale. Visual features, such as edges and corners, are carried by high-order statistics (also known as multipoint correlations or phase correlations). Analysis of discrimination of "isodipole" textures, which isolate specific kinds of high-order statistics, demonstrates visual sensitivity to these statistics but stops short of analyzing the underlying computations. Here we develop and apply a new psychophysical paradigm that gives insight into these mechanisms. Method. We focus on a canonical isodipole texture, the "even" texture -- in which any 2x2 subarray contains an even number of black and white checks. Stimuli consisted of an array of 14 texture-disks, each containing a sample of a texture drawn from the continuum from random to the pure even texture. Subjects (N=6, 1500 trials) were asked to mouse-click the centroid of the array (presented for 300 ms then masked), weighting the disks according to the evenness of the texture they contained. Performance was modeled by assuming that (1) subjects deploy a "neural filter" to derive a spatial activation map from the stimulus; (2) filter output at stimulus location (x,y) depends on the pattern within the 3x3 subblock of texture centered at (x,y); (3) the subject's estimate of the centroid is the centroid of the filter activation-map produced by the stimulus, plus response noise. Results. Subblocks whose 9 checks were all one polarity produced the highest filter-activation Asolid; subblocks comprising a single 3-check bar and six checks of opposite polarity produced an activation roughly equal to Asolid/2. All other subblocks produced activations less than Asolid/4. A 2-pass procedure showed this model accounted for nearly all of the explainable variance. Conclusions. Visual mechanisms that are sensitive to "even-ness" do not compute four-point correlations per se, but rather are activated selectively by the solid blocks and bars that occur more frequently in the even texture.

Meeting abstract presented at VSS 2016

×
×

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.

×