Sensitivity to Monocular Occlusions in Stereoscopic Imagery: Implications for S3D Content Creation, Distribution and Exhibition

Since S3D requires two views of a scene, one for each eye, transformations such as reseating, 2D to S3D conversion, synthesis of multiview displays, coding and ADAT communications efficiency require generation of new views from 2D images. One of the main challenges to this process is the identification and treatment of monocularly occluded regions. In natural environments, monocular occlusions occur whenever objects are partially obstructed by other objects in a scene, giving rise to a region that is visible to only one eye. Experiments have shown that these regions influence depth percepts. Importantly, if monocular occlusion regions are presented with texture that is inconsistent with the surrounding regions, or with inappropriate geometry, depth is degraded. This paper will review the geometric basis of monocular occlusions and their role in natural depth perception. The analysis will be framed in the context of the reconstruction of novel and appropriate viewpoints from sequences of 2D images from one or more vantage points.

Published

2010-07

Content type

Original Research

Keywords

S3D, monocular occlusions, 2D to 3D conversion, stereoscopic depth perception

DOI

10.5594/M001406

ISBN

978-1-61482-950-8