August 2009
Volume 9, Issue 8
Vision Sciences Society Annual Meeting Abstract  |   August 2009
Boundary segmentation of naturalistic textures: Roles of sparseness and local phase structure
Author Affiliations
  • Elizabeth Arsenault
    McGill Vision Research, and Department of Psychology, McGill University
  • Ahmad Yoonessi
    McGill Vision Research, and Department of Neuroscience, McGill University
  • Curtis Baker
    McGill Vision Research
Journal of Vision August 2009, Vol.9, 1042. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Elizabeth Arsenault, Ahmad Yoonessi, Curtis Baker; Boundary segmentation of naturalistic textures: Roles of sparseness and local phase structure. Journal of Vision 2009;9(8):1042.

      Download citation file:

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

  • Supplements

Some texture boundaries are easier to segment than others, and characterizing these discrepancies is an important step in understanding the neural mechanisms of texture segmentation. Previously (Baker et al., VSS 2008), we demonstrated that removing the higher-order statistics from natural textures decreases contrast boundary segmentation thresholds. We also observed that the amount of “stuff” in a texture (sparseness) appears to predict the degree to which this is the case.

To examine this issue in greater detail, we created synthetic textures that mimic the properties of natural textures. Our micropatterns were Gaussian-windowed broadband edges, created as sums of phase-aligned sine waves, which provided local edge structure and a local spectral slope of 1/f. According to the method in Kingdom et al. (Vision Research 2001), we synthesized a global spectral slope of 1/f by varying the size distributions of these micropatterns. This allowed us to (1) remove all higher-order statistics by phase-scrambling the texture, or (2) remove local phase alignments by phase-scrambling the micropatterns, while (3) varying texture sparseness by changing the number of micropatterns.

We contrast-modulated the texture to create a boundary. Participants made forced-choice judgments of boundary orientation (left- vs. right-oblique). We obtained modulation-depth thresholds using constant stimuli for intact, globally scrambled, and locally scrambled textures at a series of micropattern densities.

For sparse textures global phase-scrambling considerably reduces modulation-depth thresholds, consistent with our previous findings in natural textures. As the micropattern density is increased, the effect of global phase scrambling is progressively reduced. We observe no effect of local phase scrambling at any density level.

These results suggest that natural boundary segmentation is impeded by the sparseness of textures but not by the local edge structure of the elements that comprise them.

Arsenault, E. Yoonessi, A. Baker, C. (2009). Boundary segmentation of naturalistic textures: Roles of sparseness and local phase structure [Abstract]. Journal of Vision, 9(8):1042, 1042a,, doi:10.1167/9.8.1042. [CrossRef]
 Supported by Canadian NSERC grant #OPG0001978 to C.B.

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.