Adaptive-Scale Robust Estimator using Distribution Model Fitting Thanh Trung Ngo 1 , Hajime Nagahara 1 , Ryusuke Sagawa 1 , Yasuhiro Mukaigawa 1 , Masahiko Yachida 2 , and Yasushi Yagi 1 1 Osaka University 2 Osaka Institute of Technology Abstract. We propose a new robust estimator for parameter estimation in highly noisy data with multiple structures and without prior informa- tion 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 dis- tribution 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 es- timated 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. 1 Introduction Robust parameter estimation is fundamental research in the fields of statistics and computer vision. It can be applied in many estimation problems, such as extracting geometric models in intensity images and range images, estimating motion between consecutive image frames in a video sequence, matching images to find their similarity, and so on. In these problems, the data contains explana- tory data, which also includes leverage elements, and a large number of outliers. The data may also contain several structures, such as various lines or planes that appear in pictures or range images of a building. Therefore, the common requirements for a modern robust estimator in computer vision are: robustness to various high outlier rates (high breakdown point [1]), ability to work with multi-structural data and good detection of inliers. In this paper, we present a new robust estimator that has a high breakdown point, can work with multi-structural data and estimates the correct inlier scale. Our method relies on a novel inlier scale estimator and a density-based objective function. The proposed inlier scale estimator finds the most Gaussian-like parti- tion globally in the residual distribution of a putative solution. This is the main