Efficient Multiview Stereo by Random-Search and Propagation

We present an efficient multi-view 3D reconstruction method based on randomization and propagation scheme. Our method progressively refines 3D point estimates by ran-domly perturbing the initial guess of 3D points and propa-gates photo-consistent ones to their neighbors. In contrast to previous refinement methods that perform local optimiza-tion for a better photo-consistency, our randomization ap-proach takes lucky matchings for reducing the computa-tional complexity. Experiments show favorable efficiency of the proposed method with the accuracy that is close to the state-of-the-art methods. 1