May 2008
Volume 8, Issue 6
Free
Vision Sciences Society Annual Meeting Abstract  |   May 2008
Knowing which channel is relevant does not improve performance in texture segmentation
Author Affiliations
  • Nicolaas Prins
    Department of Psychology, University of Mississippi
Journal of Vision May 2008, Vol.8, 351. doi:10.1167/8.6.351
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to Subscribers Only
      Sign In or Create an Account ×
    • Get Citation

      Nicolaas Prins; Knowing which channel is relevant does not improve performance in texture segmentation. Journal of Vision 2008;8(6):351. doi: 10.1167/8.6.351.

      Download citation file:


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

      ×
  • Supplements
Abstract

Variations in either the orientation, spatial frequency, or contrast (or a combination thereof) in a visual texture all involve contrast variations within narrow orientation/spatial frequency channels, even in the absence of a variation in overall contrast. Accumulating evidence indicates that such variations in image statistics are detected by so-called Filter-Rectify-Filter (FRF) mechanisms which detect contrast variations within narrow orientation and spatial frequency bands. Much research has concentrated on determining whether the visual system also has available FRF mechanisms which combine information across first-order channels. Here, textures are created which contain contrast variations in two orthogonal orientation channels. The textures either contain contrast only in the relevant (i.e., contrast-modulated) channels or contain contrast in irrelevant, unmodulated channels also. Performance in both conditions is described remarkably well by a model which assumes that performance is determined by probability summation between three mechanisms: two standard FRF mechanisms (each selective for one of the two modulated channels) and an FRF mechanism which linearly combines information across first-order channels. When the texture contains contrast only in the relevant channels, performance is dominated by the FRF mechanism which combines information across orientation channels. However, when contrast in irrelevant channels is present, a mechanism which combines information across all channels would be less efficient and performance is dominated by the standard FRF mechanisms. In a second experiment the relevant information was either contained consistently within the same orientation channels on each trial or varied randomly between orientation channels. Performance did not differ between these two conditions, suggesting that standard FRF mechanisms are not labeled with respect to the orientation of the first-order channel that serves as their front-end input.

Prins, N. (2008). Knowing which channel is relevant does not improve performance in texture segmentation [Abstract]. Journal of Vision, 8(6):351, 351a, http://journalofvision.org/8/6/351/, doi:10.1167/8.6.351. [CrossRef]
×
×

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.

×