Level set stereo for cooperative grouping with occlusion

Jialiang Wang, Todd E. Zickler

Abstract: Localizing stereo boundaries is difficult because matching cues are absent in the adjacent occluded regions. We introduce an energy and level-set optimizer that improves boundaries by encoding the geometry of occlusions in stereo images that are rectified: the spatial extent of an occlusion must equal the amplitude of the disparity jump that causes it. Focusing on two-layer, figure-ground scenes, we implement the optimizer cooperatively, using messages that pass predominantly between parents and children in an undecimated hierarchy of multi-scale image patches. In a collection of figure-ground scenes from Middlebury and Falling Things stereo datasets, the model provides more accurate boundaries than previous occlusion-handling stereo techniques.