A Bayesian treatment of the stereo correspondence problem using half-occluded regions

Abstract

A half-occluded region in a stereo pair pixels in one image representing points in is a set of svace visible to that camera or eye only, and not to ihe other. These occur typically as parts of the background immediately to the left and right sides of nearby occluding objects, and are present in most natural scenes. Previous approaches to stereo either ignored these unmatchable points or attempted to weed them out in a second pass. Our algorithm incorporates them from the start as a strong clue to depth discontinuities. Psychophysical evidence suggests that the human visual system also exploits these clues. We start by deriving a measure f o r goodness of fit and a prior based on a simplified model of objects in space, which leads to an energy functional depending both on the depth as measured from a central "cyclopean" eye and on the regions of points occluded from the left and right e y e perspectives. We minimize this using dynamic programming along epipolar lines followed by annealing in both dimensions. Experiments indicate that this method is very effective even in difficult scenes.

Local Feature Selection and Global Energy Optimization in Stereo

“…It has been shown that the detection of occlusions is especially important in human stereopsis (Anderson, 1994;Nakayama & Shimojo, 1990). Occlusions have also been modelled in artificial vision systems (Belhumeur & Mumford, 1992;Geiger, Ladendorf, & Yuille, 1995). Any effect on the images due to the shape and configuration of the 3D surface can be modelled in the image formation model.…”

“…There is now abundant psychophysical evidence (Anderson, 1994;Gillam & Borsting, 1988;Nakayama & Shimojo, 1990) that the human visual system does take advantage of the detection of occluded regions to obtain depth information. The earliest attempts to model occlusions and its relation to discontinuities (Belhumeur & Mumford, 1992;Geiger, Ladendorf, & Yuille, 1995) had a limitation that they restrict the optimization function to account only for interactions along the epipolar lines. Another aspect of the stereo geometry is the interdependence between epipolar lines.…”

Disparity-space images and large occlusion stereo

“…Belhumeur has considered occlusion in several papers. In [5], Belhumeur and Mumford point out that occluded regions, not just occlusion boundaries, must be identied and incorporated into matching. Using this observation and Bayesian reasoning, an energy functional is derived using using pixel intensity as the matching feature and dynamic programming is used to nd the minimal-energy solution.…”

“…In binocular imagery, w e encounter occlusion times two. Stereo images contain occlusion edges that are found in monocular views and occluded r e gions that are unique to a stereo pair [5]. Occluded regions are spatially coherent groups of pixels that can be seen in one image of a stereo pair but not in the other.…”

Stereo-Based Dense Disparity Estimation for Lunar Surface

“…According to the calculation model, global method can be divided into the method based on Dynamic Programming (DP) [8] and the method based on Markov Random Field (MRF) [9,10]. The method based on DP method adopts scanning line optimization [11,12] and minimize the objective function through the calculation of the minimum cost path between two scanning lines. The method based on MRF includes Simulated Annealing (SA) [13], Highest Confidence First (HCF) [14], Graph Cut algorithm (GC) [15,16] and Belief Propagation (BP) [17,18].…”