Purchase this article with an account.
Guy Ben-Yosef, Ohad Ben-Shahar; Curvature-based segregation for multi-oriented textures. Journal of Vision 2009;9(8):927. doi: 10.1167/9.8.927.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Texture perception and texture segregation are traditionally linked to the notion of feature gradient (e.g., Nothdurft, 1985,1991; Landy&Bergen, 1991; Mussap&Levi ,1999). Previously, we argued (Ben-Shahar, 2006) that at least for Orientation-Defined Textures (ODTs) this link is lacking since smoothly-varying ODTs exhibit salient perceptual singularities despite having no outstanding orientation contrast. Instead, these perceptual singularities can be predicted by a perceptual measure that depends upon two orientation curvatures, one tangential and one normal, that come about naturally from considering the intrinsic differential geometry of the ODT. This measure can be computed in a biologically plausible fashion using known mechanisms in the primary visual cortex (Ben-Yosef&Ben-Shahar, 2008), thus providing additional evidence for the computation of curvatures very early in the visual process.
In this work we begin to generalize our previous results by presenting multi-oriented textures — orientation-defined textures with more than one dominant orientation at each point. Since in the limit any texture or natural image can be represented as a superposition of oriented components, possibly at different frequencies, these multi-oriented patterns constitute a critical first step towards generalizing curvature-based segregation to general textures. We therefore propose that a curvature-based model for segregation of multi-oriented textures must decompose the texture into a set of smoothly-varying orientation maps, each of which gives rise to its own perceptual singularities defined by the perceptual singularly measure discussed above. We present biologically plausible models for achieving this task and discuss further implications to general texture segregation.
This PDF is available to Subscribers Only