Binocular half-occlusions arise as a foreground surface obscures a portion of a background surface to one of the two images in a binocular pair. Half-occlusions occur at the projections of surface boundaries; therefore, they are a potential source of information in delineating 3D objects. In this paper, a novel integration of two sources of information (constraints) is proposed for binocular half-occlusion detection. The first constraint comes as an analytic relationship between the change in disparity across a half-occluded region and region width. The second constraint comes in terms of expected values of binocular match scores in the vicinity of half-occlusions. To allow for uncertainty in inference, these two constraints have been combined in a Bayesian formulation. A corresponding algorithm for half-occlusion detection, given maps of binocular disparity and match score has been implemented in software and empirically evaluated on seven standard stereo pairs. For cases where direct comparisons are possible, the developed method outperforms previous approaches to half-occlusion detection.

Download paper in PDF format.

The documents distributed by this server have been provided by the contributing authors as a means to ensure timely dissemination of scholarly and technical work on a noncommercial basis. Copyright and all rights therein are maintained by the authors or by other copyright holders, notwithstanding that they have offered their works here electronically. It is understood that all persons copying this information will adhere to the terms and constraints invoked by each author's copyright. These works may not be reposted without the explicit permission of the copyright holder.