Abstract
Visual completion, the perceptual process of completing visual information missing due to occlusion, is a fundamental organizational process that facilitates much of higher level vision. The general problem of contour completion is typically divided into the grouping problem and the shape interpolation problem. The latter deals with the ability of the visual system to complete the gaps in occluded objects and by that to facilitate recognition and other higher level visual tasks. The grouping problem, on the other hand, deals with the process that pairs specific inducers (the points in the visual stimulus where the occluder meets the occluded object) between which completion occurs. While previous computational work has addressed mostly the shape problem, perceptual work has addressed both issues. Even so, perceptual studies on the grouping problem have been restricted to the conditions by under which a given pair of inducers is indeed grouped to induce a completed contour, rather than to studying how inducers are grouped in multi-inducer scenarios. Here we investigate the grouping problem in stimuli with two pairs of inducers, exploring psychophysically the stimuli features that lead to the selection of a specific grouping decision from the set of all possible groupings. We employ a dot localization paradigm (Guttman and Kellman, 2004) and report grouping results based on the independent parameters of the geometric properties of the shape, including inducer distance, orientation difference, curvature, etc... Using these findings we developed a model that quantifies inducer grouping during visual completion in a way that extends the popular relatability theory (Kellman and Shipley, 1991) and readily allows the future incorporation of top-down factors like familiarity and attention.
Meeting abstract presented at VSS 2015