Purchase this article with an account.
Yi-Zhong Wang, Robert F. Hess; Integration of local orientational and positional contour features in shape discrimination. Journal of Vision 2003;3(9):658. doi: 10.1167/3.9.658.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Purpose: The detection of radial deformation of circular shape may involve the pooling of both local orientational and positional contour features. Here we investigated the independent grouping mechanisms for these two features. Methods: Stimuli were Gabor patch-sampled radial frequency (GSRF) patterns. Patch carrier orientation was tangential to the deformed circle at the placement position. With the change of modulation amplitude, patches placed at zero-crossings of the sinusoidal radial modulation had only orientational change from circularity but no positional alteration (orientational GSRF); patches placed at the peaks and troughs of the modulation had only positional perturbations from circularity but no orientation change (positional GSRF). The SD of Gaussian envelope was 0.14 deg; the carrier frequency was 3 cpd; the mean radius was 1 deg and the radial frequency was 4 cyc/360 deg, resulting in 8 patches in a GSRF pattern. The number of modulated patches was 1, 4 or 8. A temporal 2AFC staircase paradigm was employed. Thresholds were estimated by a maximum likelihood fitting procedure. Results: The mean thresholds of 3 normal subjects for detecting the modulation of 1, 4 and 8 elements in orientational GSRF patterns were 2.72±0.44(SD)%, 1.16±0.45% and 0.67±0.25%, respectively. The corresponding thresholds for positional GSRF patterns were 2.25±1.0(SD)%, 1.47±0.6% and 1.83±1.0%, respectively. While the pooling of local positional perturbations reached a plateau at 4 elements of modulation, the integration of local orientation continued linearly in a log-log scale until all 8 modulated elements were pooled. The slope of orientation pooling was 0.67, more than twice that for positional pooling. Conclusions: While both orientation and position of local contour elements can encode the deformation of circular shape, the pooling of local orientation into global shape is much stronger than that of local position and involves pooling over a larger spatial extend.
This PDF is available to Subscribers Only