A novel 3D reconstruction approach

3D model reconstruction of a real scene is an important and active research topic in computer vision. In this pa,per, we propose a novel 3D reconstruction algorithm by combining shape from silhouette with stereo. We first construct a visual hull of the object from multi-view silhouette images. Then we accomplish pairwise stereo matching for shape refinement using the best viewable images. Based on the reduced correspon¬dence searching range constrained by contact points and bounding edges, visual hull is significantly improved even if the number of cameras is limited. Results show that our approach performs well for both synthetic data and real scene images.