Adaptively Up-Sampling Point-Sampled Models

Adaptively up-sampling of point-sampled models is one of key technologies to build multi-resolution point-based surfaces. In this paper, we propose an up-sampling algorithm for point-models. Our algorithm first defines a smooth manifold patch for each point in model based on a local projecting procedure, defined by the famous moving least squares (MLS) method. A valid up-sampling region is calculated for each point via constructing local Voronoi diagram of the point with its neighbors. Then up-sample the region following a grid which constructed on the local plane and commonly controlled by the local curvature of the surface and a given sampling control function. Collecting the up-sampling points, a dense multiresolution point-model can be obtained.