Improved random sample consensus algorithm with near point consistency

Random sample consensus (RANSAC) algorithm is one of the most widely used approaches in the field of computer multi-view vision. To improve the efficiency of initial matching with high noise, a new filter algorithm is proposed based on near point consistency. The new method improves the inliers ratio of initial matching set and reduces the iteration times when using RANSAC to find parameters. The experiment on wide-based pictures demonstrates that the new algorithm achieves a significant increase in speed and almost the same result compared with the classical RANSAC algorithm.