The Statistical Relationship between Depth, Visual Cues, and Human Perception

Abstract This chapter uses the Bayesian framework to explore the information content of some underappreciated sources of depth information: the shape of the contour dividing two image regions and the pattern of blur across the retinal image. It argues that previous claims that blur is a weak depth cue providing only coarse ordinal information are incorrect. When the depth information contained in blur is represented in the Bayesian framework, it provides useful information about metric depth when combined with information from nonmetric depth cues like perspective. The conventional, geometry-based taxonomy that classifies depth cues according to the type of distance information they provide is unnecessary. By capitalizing on the statistical relationship between images and the environment to which the study's visual systems have been exposed, the probabilistic approach used in this chapter aims to yield a richer understanding of how 3D layout is perceived.