作者: Alexander Schwing , Tamir Hazan , Marc Pollefeys , Raquel Urtasun
DOI: 10.1109/CVPR.2011.5995642
关键词: Message passing 、 Computer graphics 、 Consistency (database systems) 、 Computation 、 Theoretical computer science 、 Scale (ratio) 、 Parallel algorithm 、 Graphical model 、 Computer science 、 Inference
摘要: In this paper we propose a distributed message-passing algorithm for inference in large scale graphical models. Our method can handle problems efficiently by distributing and parallelizing the computation memory requirements. The convergence optimality guarantees of recently developed algorithms are preserved introducing new types consistency messages, sent between computers. We demonstrate effectiveness our approach task stereo reconstruction from high-resolution imagery, show that is possible with more than 200 labels images larger 10 MPixels.