Automatic RANSAC by Likelihood Maximization

In computer vision, and particularly in 3D reconstruction from images, it is customary to be faced with regression problems contaminated by outlying data. The standard efficient method deal them the Random Sample Consensus (RANSAC) algorithm. procedure simple versatile, drawing random minimal samples data estimate parameterized candidate models ranking based on amount of compatible Such evaluation involves some threshold that separates inliers outliers. presence unknown level noise, as usual practice, desirable remove dependency this fixed an additional unknown. Among numerous variants RANSAC, few, we call 'automatic', propose approach, which changing maximization criterion consensus, naturally increasing varying threshold. An algorithm Zach Cohen (ICCV 2015) uses likelihood statistics. We present details implementation their along quantitative qualitative tests stereovision tasks: estimation homography, fundamental essential matrix.