Feature-aware Methods for Mesh and Point Cloud Processing

Abstract

Due to the recent development of 3D scanning and modelling techniques, 3D digital shapes are largely available for numerous applications such as visual effects, computational design, virtual reality, just to name a few. As a result, shape processing as a major area in computer graphics becomes more and more important. Conventional approaches manipulate a surface shape by processing individual surface points and defining distances between points in an isotropic way. However, the important anisotropic sharp features of the shape are not taken into account, making existing work unsuitable for feature-dependent problems. In this dissertation, we present three novel feature aware methods for mesh and point cloud processing that fill the literature gap, and can be utilized for potential applications including architectural design, physical simulation, and digital acquisition. In the first work, we present a novel optimization method for generating anisotropic ellipsoid packing structure on freeform surfaces with robust remeshing-based initialization. The resultant ellipsoid packing structures can also be used to generate appealing architectural designs. In the second work, we propose a novel anisotropic surface remeshing method that can efficiently eliminate obtuse angles. It relies on a simple yet efficient connectivity and geometry refinement, which can not only remove all the obtuse angles, but also preserves the original mesh connectivity as much as possible. In the third work, we present a novel point cloud denoising method based on a generalized robust metric. This allows a uniform optimization framework that emphasizes on noise removal and feature preservation adaptively, thus can avoid typical inferior cases of over-smoothing and keep sharp features.

Date of Award	21 Jul 2021
Original language	English
Awarding Institution	University of Bath
Supervisor	Yongliang Yang (Supervisor) & Peter Hall (Supervisor)