Poisson Surface Reconstruction, Symposium on Geometry Processing, 2006
Poisson Surface Reconstruction is a method for reconstructing a 3D surface from oriented point samples. The key insight is to reformulate the problem as a Poisson equation, where the goal is to find an indicator function $\tilde\chi$ whose gradient best matches the input normal field derived from the oriented points. Using locally supported basis functions, the problem reduces to solving a sparse linear system, which can be done efficiently even for large point clouds. The resulting method is robust to noise and can handle non-uniform sampling, making it one of the most popular choices for surface reconstruction tasks.
Play with the sample count and noise levels to see how they affect the reconstruction quality. The demo uses precomputed reconstructions for a fixed set of parameter combinations, so you can quickly compare results without waiting for the solver to run in real time.
Toggle between the samples, the surface, or both views together.