Abstract
In this paper we propose a simple method for the decomposition of point-sampled 3D objects into its basic components. Our approach is based on recent methods for fuzzy clustering and hierarchical decomposition of 3D meshes where we use instead a k-nearest neighbor graph of the point cloud. Our approach proceeds in two steps: We first encode the geometric properties of the 3D shape into an inter-surfel distance matrix. The distance between two surfels takes into account the geodesic distance the angular distance to measure the shape convexity and the surface variation. Then we apply a clustering algorithm on the distance matrix to extract the different components of the input surfels. We demonstrate the efficiency of the proposed approach on a collection of point sampled 3D objects.In this paper we propose a simple method for the decomposition of point-sampled 3D objects into its basic components. Our approach is based on recent methods for fuzzy clustering and hierarchical decomposition of 3D meshes, where we use instead a k-nearest neighbor graph of the point cloud. Our approach proceeds in two steps: We first encode the geometric properties of the 3D shape into an inter-surfel distance matrix. The distance between two surfels takes into account the geodesic distance, the angular distance to measure the shape convexity and the surface variation. Then, we apply a clustering algorithm on the distance matrix to extract the different components of the input surfels. We demonstrate the efficiency of the proposed approach on a collection of point sampled 3D objects.