Space-Partitioning RANSAC

A new algorithm is proposed to accelerate the RANSAC model quality calculations. The method based on partitioning joint correspondence space, e.g., 2D-2D point correspondences, into a pair of regular grids. grid cells are mapped by minimal sample models, estimated within RANSAC, reject correspondences that inconsistent with parameters early. technique general. It works arbitrary transformations even if set, as fundamental matrix maps epipolar lines. tested thousands image pairs from publicly available datasets and essential matrix, homography radially distorted estimation. On average, space reduces run-time 41% provably no deterioration in accuracy. When combined SPRT, drops its 30%. can be straightforwardly plugged any state-of-the-art framework. code at https://github.com/danini/graph-cut-ransac .