Adaptive estimation of inlier and outlier threshold

Abstract

One of the main problems relating RANSAC estimation is to determine the inlier threshold adaptively depending on the variance of inliers. In this paper, we propose a novel method that estimates the inlier threshold adaptively from the observations, giving a threshold-free RANSAC. A minimum assumption of our method is that the lower bound of inlier ratio is known in advance and the variance of inliers follows Gaussian. We also describe a simple motion flow tracker as an application of the proposed method. In the experiment we show the effectiveness of the proposed method by comparing tracking performance with and without adaptive estimation of inlier threshold.