Surface Reconstruction from Unorganized Point Clouds

This research work proposes a new surface reconstruction algorithm for 3D unorganized point clouds based on the concept of contour volution on 2D manifolds. The algorithm starts from a simple seed (e.g. a triangle) that can be automatically initialized, and always evolves its boundary contour outwards along its tangent direction suggested by the underlying point clouds data. The new algorithm can create high quality semi-regular meshes that are either uniform or adaptive to the local shape curvature. Experiment results on both real and synthetic datasets are shown to demonstrate the robustness and effectiveness of the new algorithm.