A Robust Method for Estimating the Fundamental Matrix

In this paper, we propose a robust method to estimate the fundamental matrix in the presence of outliers. The new method uses random minimum subsets as a search engine to find inliers. The fundamental matrix is computed from a minimum subset and subsequently evaluated over the entire data set by means of the same measure, namely minimization of 2D reprojection error. A mixture model of Gaussian and Uniform distributions is used to describe the image errors. An iterative algorithm is developed for estimating the outlier percentage and noise level in the mixture model. Simulation results are provided to illustrate the performance of the proposed method.