Spectral Surface Reconstruction

This project aims to build a package for reconstruction watertight surfaces from un-oriented point sets. The first step is to estimate un-oriented normals to the inferred surface by combining Voronoi diagram and Principal Component Analysis. It results in a tensor field which encodes both the normal direction and the confidence in the approxiamtion. The second step is to estimate a implicit function such that its gradient is best aligned with the principal axes of the tensor field encoding the normal directions. This estimation could be viewed as a generalized eigenvalue problem, which can be solved by the power iteration.