Abstract
Monocular and binocular recovery of 3D symmetrical and near-symmetrical shapes Vijai Jayadevan, Tadamasa Sawada, Edward Delp and Zygmunt Pizlo Our prior work demonstrated that perception of 3D symmetrical shapes can be explained by a cost function that combines a priori constraints such as symmetry and compactness with binocular depth-order information. In this study we show that a cost function containing these terms can explain perception of near-symmetrical shapes, as well. In the experiment, performed in a virtual reality environment using Oculus Rift, the subject adjusted 3 parameters of a rotating 3D shape to match the percept of a stationary reference shape. The reference shape was either mirror-symmetrical or asymmetrical. The asymmetrical shapes were produced by applying an affine transformation to the symmetrical shape. Ninety objects of varying degree of symmetry and compactness were used in the experiment. Three subjects were tested. The subject adjusted 3 parameters representing all 3D affine transformations of the 3D shape that keep the 2D cyclopean orthographic image unchanged. These parameters included a uniform stretch and compression along the depth direction. Our results show that binocular perception of 3D symmetrical shapes is veridical. In particular, there is no systematic under- or overestimation of depth. Binocular perception of near-symmetrical shapes is less veridical – the percept is biased by the symmetry and compactness constraints. Monocular perception of symmetrical and near symmetrical shapes can be explained by symmetry and compactness priors. Individual differences are minimal, if present at all, in monocular viewing. Binocular viewing shows appreciable individual differences in the case of asymmetrical shapes. These individual differences correspond to the relative weights of the symmetry, compactness and binocular depth-order terms in the cost function.
Meeting abstract presented at VSS 2018