How We See the World in Depth

This chapter explains how 3D vision and figure-ground perception occur in our brains. It shows how the 2D boundary and surface processes that are described in earlier chapters naturally generalize to 3D via both the FACADE (Form-And-Color-And-DEpth) theory of 3D vision and figure-ground perception, and the 3D LAMINART model that generalizes the laminar cortical circuits of Chapter 10 to 3D and naturally embodies and generalizes FACADE. Contrast-specific binocular fusion and contrast-invariant boundary formation are explained in terms of identified cells in specific layers of cortical areas V1 and V2. The correspondence problem is solved using a disparity filter that eliminates false binocular matches in layer 2/3 of V2, while it chooses the 3D binocular boundary grouping that is best supported by scenic cues. The critical role of monocular boundary information in figure-ground perception is explained and used to simulate DaVinci stereopsis percepts, along with surface-to-boundary surface contour signals and a fixation plane bias due to life-long experiences with fixated scenic features. Simulated data include the Venetian blind effect, Panum’s limiting case, dichoptic masking, 3D Craik-O’Brien-Cornsweet effect, Julesz random dot stereograms, 3D percepts of 2D pictures of shaded ellipses and discrete textures, simultaneous fusion and rivalry percepts when viewing Kulikowski and Kaufman stereograms, stimulus rivalry and eye rivalry, and bistable percepts of slanted surfaces, including the Necker cube. The size-disparity correlation enables signals from multiple scales to cooperate and compete to generate boundary representations at multiple depths. 3D percepts of natural scenes from stereograms are also simulated with these circuits.