The impact of monocular self-occlusions on depth perception

Abstract

Depth perceived from two-object arrangement half-occlusions has been shown to depend on the occluded regions size and texture properties. However, monocular regions also result from single-object self-occlusions where the monocular region sizes and texture properties could influence its perceived shape. In a series of experiments, I assessed sensitivity to monocular texture changes, perceived object coherence, and occlusion size under geometrically valid and invalid conditions. Monocular texture changes had little impact on perceived depth when binocular disparity was present in the stimulus suggesting that disparity is extrapolated across monocular regions to determine object shape. There was considerable tolerance to invalid occlusion arrangements when monocular regions were perceived as part of a coherent object. Furthermore, quantitative depth from self-occlusions do not follow minimum depth constraint predictions. My experiments show that interpretation of self-occlusions depends on their context; heuristics used to extract depth information from monocular regions are contingent on foreground/background segmentation.