Shape from silhouette using topology-adaptive mesh deformation

We present a computationally efficient and robust shape from silhouette method based on topology-adaptive mesh deformation, which can produce accurate, smooth, and topologically consistent 3D mesh models of complex real objects. The deformation scheme is based on the conventional snake model coupled with local mesh transform operations that control the resolution and uniformity of the deformable mesh. Based on minimum and maximum edge length constraints imposed on the mesh, we describe a fast collision detection method which is crucial for computational efficiency of the reconstruction process. The topology of the deformable mesh, which is initially zero genus, can be modified whenever necessary by merging operations in a controlled and robust manner by exploiting the topology information available in the silhouette images. The performance of the proposed shape from silhouette technique is demonstrated on several real objects.