An Algorithm for Image Match Propagation
Rong Liu, Liang He, LU Kai-yun · 2012
Stereo matching is an important part of 3D reconstruction. This paper proposes an improved match propagation algorithm. Firstly, the feature points are extracted and matched by SIFT algorithm then they are used as seed points for the image match propagation in neighborhood. Secondly, these dense correspondences are re-sampled to quasi-dense correspondences. Finally, the mismatched points and outliers are eliminated by epipolar constraint. Experiments show that this improved algorithm can be applied in the similar structure of images with great change of depth.