Multi-view Superpixel Stereo in Man-made environments

Google Tech Talks April, 7 2008 ABSTRACT Man-made environments possess many regularities which can be efficiently exploited for 3D dense reconstruction from multiple widely separated views. We present an approach utilizing properties as piecewise planarity and restricted number of plane orientations to suppress the ambiguities causing failures of standard dense stereo methods. We formulate the problem of 3D reconstruction as one of MRF inference over an image presegmented into superpixels. Using this representation, we propose novel robust cost measures which handle favorably problematic scenarios containing many repetitive structures and textureless or low textured regions. We demonstrate our approach on several low textured wide-baseline scenes demonstrating superior performance compared to previously proposed methods. Speaker: B. Micusik, J. Kosecka