Multi-view Reconstruction Using Band Graph-Cuts

This paper proposes a new multi-view reconstruction algorithm by combining the visual hull and multi-view stereo matching based on the band graph-cuts technique.The reconstructed surface is made consistent with the dense point cloud obtained from the multi-view stereo matching through the minimization of the functional that is constrained by the topology of the built visual hull.The globally optimal solution can be achieved with the minimum cut of a band weighted graph meanwhile the overall performance is greatly improved.Experimental results show our method is able to recover surface concavities and watertight surface.Compared with conventional graph-cuts approaches,more than three folds acceleration is gained with our approach.