Belief propagation on Riemannian manifold for stereo matching

Stereo matching has been one of the most active research areas in computer vision for decades. Many methods, ranging from similarity measures to local or global matching cost optimization algorithms, have been proposed. As we known, stereo matching can be formulated under the framework of Markov random field (MRF), and the global optimization in stereo matching can be approximated by inference procedure. There are many exact or approximate inference algorithms, among which belief propagation is one of the most effective. In this paper, by combining Riemannian metric based similarity measure with the belief propagation algorithm, we propose a global optimization method for stereo matching, namely belief propagation on Riemannian manifold (BPRM). Experiments on benchmark dataset demonstrate the encouraging performance of our method.