Robust Surface Consolidation of Scanned Thick Point Clouds

This paper proposes a consolidation method for scanned point clouds that are usually corrupted by noises, outliers, and thickness. At the beginning, we construct neighborhood of a point based on shared nearest neighbor relationship. Then, the points with few number of neighbors are regarded as outliers and removed. After that, we propose a feature-aware projection operator to thin the thick point clouds by considering spatial distances, normal diversifications, and the squash directions of thick point clouds. Experiment results of scanned point clouds show that our method can consolidate the thick point clouds while preserving sharp features and geometry details.