Silhouette-Based variational methods for single view reconstruction

摘要: We explore the 3D reconstruction of objects from a single view within an interactive framework by using silhouette information. In order to deal with highly ill-posed nature problem we propose two different priors: shape and volume prior cast them into variational formulation. For both priors show that corresponding relaxed optimization is convex. This leads unique solutions which are independent initialization either globally optimal (shape prior) or can be shown lie bounds solution (volume prior). analyze properties proposed regard results as well their impact on minimization problem. By employing implicit volumetric representation our reconstructions enjoy complete topological freedom. Being parameter-based, tool allows for intuitive easy use modeling result.