Investigating the effect of texture edges in figure-ground segregation using psychophysical and eye tracking experiments

Although it happens infrequently in the natural world, the human visual system is able to perceive objects defined solely by a difference in texture i.e. with no accompanying change in colour or luminance. However, studies on texture perception have frequently used figure-ground patterns with abrupt texture variations, with few studies using patterns with smooth texture variations. The work presented in this thesis considers the contribution of the edge and centre regions of a texture figure to orientation-based texture segregation. To this end, we used psychophysical (Chapters 4 – 6) and eye tracking techniques (Chapters 8 – 10) to investigate texture detection and segmentation. For the psychophysics experiments, the primary goal was to investigate if either an edge-based or region-based mechanism can account for both smooth and abrupt texture variations. We first examined this in Chapter 4, where we studied the effects of texture edges on figure-ground segregation. Lower thresholds were found when the texture figure had orientation contrast information at the edge and centre of the figure. Data modeling supports the notion that texture segregation involves a large-scale second-order texture filter i.e. akin to a region-based mechanism, but where information is extracted over a large albeit fixed-size region. Various studies were also conducted to determine what aspects of a texture figure would change the size of the second-stage filter. The size and aspect ratio of the figure were manipulated (Chapter 5), and also the spatial frequency of the texture pattern, age of the participants, and the viewing distance (Chapter 6). We found that higher spatial frequencies resulted in larger integration regions i.e. feeds into large second-stage filters, but age, viewing distance, figure size and aspect ratio did not influence the size of the integration region. For the eye tracking studies, the general aim was to investigate what information of a texture target is extracted in order to produce signals for eye movement control. We measured eye movements made by participants while they searched for a texture figure embedded in a background. We found that irrespective of the types of orientation profiles, area-normalized data were that the centre region of a figure was looked at most often, and for longer durations. However, figures with information of orientation contrast at both the edge and centre of figure were easier to localise (Chapter 7), and produced the highest level of saliency in attracting eye movements (Chapter 10). In Chapter 9, we demonstrate that the visual system is also able to efficiently segregate a texture figure from the ground to accurately plan a saccade to the target figure, and these saccades are planned based on the representation of the whole figure shape as opposed to local salient regions. More specifically, saccades were directed to the centre of gravity of the target, with some degree of undershoot. Finally, the similar size of the integration region for the eye tracking (Chapter 10) and psychophysics experiments implies that the saccadic system receives input from the mechanism that segregates figure-ground texture stimuli