Measuring the self-consistency of stereo algorithms

Abstract

A new approach to characterizing the performance of point- correspondence algorithms is presented. Instead of relying on any“gro- und truth’, it uses the self-consistency of the outputs of an algorithm independently applied to different sets of views of a static scene. It al- lows one to evaluate algorithms for a given class of scenes, as well as to estimate the accuracy of every element of the output of the algorithm for a given set of views. Experiments to demonstrate the usefulness of the methodology are presented.