UCLA

Perception of three-dimensional (3D) structure from the two-dimensional (2D) pattern of image velocities on the retina remains a fundamental issue in vision science. 3D reconstruction is complicated by ambiguous output of local motion estimates derived from direction-selective cells in early visual cortex, as only motion orthogonal to the contour is perceived. Previous research suggests the global integration performed by the human visual system adopts constraints, such as a preference for minimal amount of shape change (Wallach & O’Connell, 1953; Jansson & Johansson, 1973; Ullman, 1979) or slowest and smoothest velocity field (Hildreth, 1984; Yuille & Grzywacz, 1989; Weiss, Simoncelli, & Adelson, 2002), reflecting systematic statistical regularities in the environment to restrict the potential 3D interpretations.

In the current study, we will employ stereokinetic phenomena, which are 2D images that result in the perception of non-veridical 2D and 3D percepts when rotated about an axis perpendicular to the image plane, to probe the constraints underlying the integration scheme and determine how motion information can allow depth. Previous research conducted in the laboratory has suggested that the visual system applies preferences for minimum motion and minimal deformation (i.e., maximal rigidity) when viewing stereokinetic stimuli (Rokers, Yuille, & Liu, 2006; Xing & Liu, 2018). In order to facilitate the development of computational models that test whether the visual system prefers a 3D object that results in the minimal amount of change and slowest velocity field could be generalized to other stereokinetic stimuli, we have developed various measurement methods to rigorously quantify the perceived depth of the stereokinetic cylinder. Across both experiments measuring observers’ perceived depth of the stimulus, the length of the illusory cylinder was constrained by a preference for slow motion and maximal rigidity. To our knowledge, we are the first to quantify the perceived length of the stereokinetic cylinder.