Computational Study of Depth Perception for an Ambiguous Image Region: How Can We Estimate the Depth of Black or White Paper?

Abstract

We propose a new computational model that accounts for human perception of depth for “ambiguous regions,” in which no information exists to estimate binocular disparity as seen in black and white papers. Random dot stereograms are widely used examples because these patterns provide sufficient information for disparity calculation. Then, a simple question confronts us: “how can we estimate the depth of non-textured images, like those on white paper?” In such non-textured regions, mathematical solutions of the spatial disparities are not unique but indefinite. We examine a mathematical description of depth estimation that is consistent with psychological experiments for non-textured images. Using computer simulation, we show that resultant depth-maps using our model based on the mathematical description above qualitatively reproduce human depth perception.