An investigation into the interactions of visual texture and shape in human vision

Determining the shape of an object plays an important role in being able to identify and interact with objects in the environment. The limits of an object can be signalled in multiple ways; both static and moving shapes usually have outlines and also textural differences. This thesis utilizes psychophysical methods to investigate how these cues are used and combined to make shape judgements. In the studies that follow, Radial Frequency (RF) patterns are employed as shape primitives. RF patterns are created by introducing a sinusoidal modulation to the radius of a circle (Wilkinson, Wilson, & Habak, 1998), and they can be easily manipulated to represent a variety of simple and complex shapes. Further, studies have demonstrated that from mid-level perceptual processes onwards low frequency RF patterns are processed globally (Loffler, Wilson, & Wilkinson, 2003), indicated by deformation discrimination thresholds falling at a rate steeper than probability summation as cycles of modulation are added. The first part of this thesis investigates whether both contours defined by luminance and by texture differences can signal global shape. We observe that the mechanism for global integration appears to be agnostic to how the boundary of a shape is defined (whether by texture- or luminance-contrast). Moreover this mechanism is sufficiently robust that integration strength is unaffected when a contour is embedded in noise. However, when used in tandem, we observed that globally processed texture and luminance boundaries make independent contributions rather than sum linearly. Modulated RF textures which sinusoidally modulate the orientation of local Gabor elements but have no intentional contiguous contours were also observed to be globally integrated when the percept of the elements flowing into closure (flowsure) was evident but failed to do so when flowsure was absent. The second part of the thesis explored the summation of such global RF textures across area. An absence of discontinuity in a texture generally denotes one object while a boundary generally denotes the border or outer-limits of an object. The section asked whether a boundary would necessarily result in a discontinuity in the global integration of a texture. It was observed that modulated RF textures did not sum across annular area even with the absence of a path. Taken with other findings, this leads to the proposal that modulated texture detectors differ from Glass pattern and RF pattern detectors. The earlier experiments demonstrated that RF patterns and modulated textures can be globally processed, albeit with different requirements (the former required specific positional information while the latter required flowsure). The final part of the thesis investigated the manner in which the visual system would treat a contour when a texture was applied to it (i.e. as a texture or as a contour). The results show that a modulated textured contour is not limited by the requirements for global integration of textures, and is instead treated as a second-order contour, i.e.; the contrast envelope is used to define the shape. It is then shown that when first- and second-order cues are presented on the same contour, the visual system utilizes them equally and they make a linear contribution to shape discrimination thresholds. Overall this thesis has looked at the manner at which shapes and visual textures interact, specifically how texture informs shape discrimination. We have shown that the mechanism that governs global processing of shape is sufficiently robust to accommodate textural noise without detriment to integration strength, and is agnostic to the manner in which the global shape is specified (i.e. texture-defined or luminance-defined border). Globally-processed textures that can be used to describe shape without the presence of boundaries were also presented and it was shown that annular regions of such texture summed probabilistically and inserting a potentially segmenting border between these regions had no impact. Additionally we observe that when a texture is applied to a contour, the visual system regards the textured-contour as a second-order contour rather than as a texture for purposes of global processing but that when first- and second-order components are combined on a contour, they contribute equally and linearly to shape discrimination. These results will facilitate the consideration of both texture and border contributions to shape discrimination in human vision.