August 2012
Volume 12, Issue 9
Vision Sciences Society Annual Meeting Abstract  |   August 2012
A model of target detectability across the visual field in naturalistic backgrounds
Author Affiliations
  • Chris Bradley
    Center for Perceptual Systems, University of Texas at Austin
  • Wilson S. Geisler
    Center for Perceptual Systems, University of Texas at Austin
Journal of Vision August 2012, Vol.12, 318. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Chris Bradley, Wilson S. Geisler; A model of target detectability across the visual field in naturalistic backgrounds. Journal of Vision 2012;12(9):318.

      Download citation file:

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

  • Supplements

Rigorous analysis of behavior and physiology in most natural tasks requires first characterizing the variation in early visual processing across the visual field. We describe a model of target detectability in uniform, naturalistic, and natural backgrounds. For detection in uniform luminance backgrounds, the model reduces to a physiological model of retinal processing that includes the optical point spread function, a sampling map of P ganglion cells, a difference-of-Gaussians model of ganglion cell processing, and a near-optimal pooling function over ganglion cell outputs. Parameters for this retinal component of the model were either fixed from anatomy and physiology or estimated by fitting the detection thresholds reported for the set of stimuli used in the ModelFest1 project. For detection in noise or natural backgrounds, the model adjusts the predicted contrast detection thresholds of the retinal processing component using a known empirical relation: the square of the threshold in white and 1/f noise backgrounds is proportional to the square of the background noise contrast plus an additive constant. This additive constant equals the square of the threshold on a uniform background, which is the prediction of the retinal processing component of the model. In the model, masking (the slope of the above masking function) depends on both the contrast power of the background that falls within the critical band of the target, and on a more broadband contrast gain control factor. The model has been implemented efficiently so that predictions can be generated rapidly for arbitrary backgrounds, target locations, and fixation locations. The model works well at predicting contrast detection thresholds across the visual field in uniform and naturalistic noise backgrounds, for targets similar to those used in the ModelFest project, but remains to be tested on natural backgrounds.

1 Watson & Ahumada (2005) Journal of Vision, 5, 717-740.

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.