A novel suface reconstruction method for noisy point cloud

摘要

Point-sampled models generated by 3D-scanning devices are invariably noisy. It is necessary to remove the inevitable noise before realizing surface reconstruction while preserving the underlying surface features in computer graphics. But the point cloud will be incomplete after denoising. In this paper, a novel and surface reconstruction method for noisy point cloud is proposed. First, the noises of the point cloud are deleted by robust ellipsoid criterion, then a surface reconstruction method based on neural network is used to reconstruct surface.