Ying He
Nanyang Technological University Singapore
Homepage
Talk title
3D Reconstruction for Unoriented Point Clouds: From Watertight Solids to Open Surfaces
Abstract
Reconstructing 3D surfaces from unoriented point clouds is a fundamental problem in digital geometry processing and 3D vision. Implicit fields provide a flexible framework, but the choice of field depends on the target geometry. For watertight surfaces, Poisson reconstruction and generalized winding numbers exploit a well-defined inside–outside distinction. I will review our work on reconstructing such surfaces without prescribed point orientations, from Iterative Poisson Surface Reconstruction (iPSR) and Diffusing Winding Gradients (DWG) to Dirichlet Winding Reconstruction (DiWR), which also adapts sampling weights and point confidence to handle uneven sampling and outliers. I will then describe recent approaches for representing and reconstructing open sheets, thin structures, and non-manifold junctions, where a global inside–outside label may be ill-defined. Together, these methods extend reconstruction from watertight solids toward more general surface geometry.
Biography
Ying He is a Professor in the College of Computing and Data Science at Nanyang Technological University, Singapore. He serves as Director of the Centre for Augmented and Virtual Reality and Programme Co-Director of the MSc in Game Design and Technology. His research spans geometric modeling, digital geometry processing, 3D deep learning, and 3D vision, with recent work focused on reconstruction from point clouds and multi-view images. He has served as General or Program Co-Chair of 10+ international conferences in geometric modeling and computer graphics. He is a Summit Co-Chair of the 4th International Geometry Summit in 2027, the first edition to be held in Asia. The summit will bring together SPM, SMI, SGP, GMP, and SIAM GD. For more information, visit https://personal.ntu.edu.sg/yhe/.