Exploiting high level scene cues in stereo reconstruction

We present a novel approach to 3D reconstruction which is inspired by the human visual system. This system uni-fies standard appearance matching and triangulation tech-niques with higher level reasoning and scene understand-ing, in order to resolve ambiguities between different inter-pretations of the scene. The types of reasoning integrated in the approach includes recognising common configura-tions of surface normals and semantic edges (e.g. convex, concave and occlusion boundaries). We also recognise the coplanar, collinear and symmetric structures which are es-pecially common in man made environments. 1