Performance Comparison of Stereo Correspondence Algorithms in Dense Image Matching

Abstract

Stereo matching is one of the most active research fields in computer vision. The aim of stereo matching is to find the corresponding points in two or more images that correspond to the same physical entity in the scene. In this paper, we deal the problem of stereo vision and precisely the dense stereo matching of images using correlation measures and other algorithms. We consider eight correlation techniques, subpixel estimation, dynamic programming and Hierarchical matching. The aim of this evaluation is to show the performance of each method in dense stereo matching images and in the 3D reconstruction. We also consider noisy image pairs with different noise level. The performance evaluation of each technique is conducted in term of: computational time, mean absolute error, mean relative error and the percentage of correct matches as well as wrong matches. A 3D reconstruction will be considered for the better methods.