Characterizing neural mechanisms underlying the extraction of orientation flows i

Li, Andrea

Abstract

The correct perception of 3-D shape enables us to functionally interact with objects in our environment. Yet how the visual system transforms retinal images into 3-D percepts is still not well understood. We have shown that in images of textured surfaces, correct perception of 3-D shape depends on the visibility of specific orientation flow patterns that, unlike other texture changes, provide a robust cue to 3-D shape. Orientation flows also play a role in the perception of specular and shaded surfaces, thus it is important that we understand how these patterns are neurally extracted. The proposed work aims to characterize the neural mechanisms that extract these patterns. There are three major goals. The first goal is to explore the role of cross-orientation suppression (COS) in the visibility of the orientation flows and shape perception. Preliminary results show that the release of a frequency-selective COS process increases the visibility of orientation flows when surfaces are slanted. Experiments are proposed to determine the orientation-selectivity of these mechanisms, and their role in slant and shape perception. Results may help explain previous findings showing improved slant discrimination at steeper slants, the efficacy of oriented textures at conveying slant, and the facilitative role of size gradients in shape perception. The second goal is to examine the relative contributions of orientation and frequency modulations in the perception of surface slant and shape. Previous observations suggest that frequency modulations enhance the shape percept;however, they can also provide incorrect cues to distance. Experiments are proposed to examine the relationship between the detection of orientation and frequency modulations and how their detection influences slant discrimination when both sources of information are available. The third goal is to investigate the spatial parameters over which local orientation information is pooled to form the requisite orientation flow patterns. Glass patterns convey 2-D structure with sparse orientation cues thought to be coded on the front end by oriented neurons in early cortical areas. Experiments are proposed to map translational Glass patterns and Gabor textures onto 3-D surfaces such that the elements are aligned along the requisite orientation flows, and to determine the minimum densities of the elements required to integrate local orientation information into requisite flow patterns that convey 3-D shape. These patterns will also be used in adaptation paradigms to tap the existence of global 3-D shape-selective neurons.

Public Health Relevance

Correct perception of 3-D shape enables us to functionally interact with objects in our environment. Yet how the brain transforms retinal images into 3-D percepts is still not understood. We investigate the neural mechanisms that enable perception of 3-D shape from orientation flow patterns.