Adaptive-scale robust estimator using distribution model fitting

We propose a new robust estimator for parameter estimation in highly noisy data with multiple structures and without prior information on the noise scale of inliers. This is a diagnostic method that uses random sampling like RANSAC, but adaptively estimates the inlier scale using a novel adaptive scale estimator. The residual distribution model of inliers is assumed known, such as a Gaussian distribution. Given a putative solution, our inlier scale estimator attempts to extract a distribution for the inliers from the distribution of all residuals. This is done by globally searching a partition of the total distribution that best fits the Gaussian distribution. Then, the density of the residuals of estimated inliers is used as the score in the objective function to evaluate the putative solution. The output of the estimator is the best solution that gives the highest score. Experiments with various simulations and real data for line fitting and fundamental matrix estimation are carried out to validate our algorithm, which performs better than several of the latest robust estimators.