Progressive Matching Based On Segmentation for 3D reconstruction
Yong Zhuang, Ma Ma, Jun Wei, Liu Liu · 2005
Dense matching plays an important role for 3D scene reconstruction. A simple rectified algorithm is given in this paper. Then the new matching algorithm starts from GCPS as seeds with rectified stereo pairs. First color segment is used to capture disparity discontinuity and the reliable seed point from GCPS which is computed by using fast ZNCC based on asymmetric window can propagate the other correspondence points according to continuity constraint in each region. The algorithm is demonstrated on standard images pairs and shows better applications on 3D reconstruction