3D Facial Reconstruction from Point Clouds with Hermite Radial Basis Functions and Multilevel Partition of Unity

• Ano de defesa: 2024
• Autor(a) principal: Tubone, Guilherme Hideo
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Faces; Funções de base radial de Hermite; Hermite radial basis functions; Multilevel partition of unity; Nuvem de pontos; Partição da unidade multinível; Point clouds; Reconstrução de superfícies; Surface reconstruction
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/55/55134/tde-02072024-085526/

Point cloud surface reconstruction is the process of transforming raw data from 3D scanners into a digital representation of an object surface. This process has applications in various fields, and as a result, many methods have been proposed to solve this problem over the years. In this work, we focus on methods that generate surfaces represented as level sets of implicit functions. Specifically, we propose a method based on Hermite Radial Basis Functions (HRBF) with Multilevel Partition of Unity (MPU). The HRBF allows a precise and detailed representation of the surface geometry, even when dealing with nonuniform data, while the MPU enables a efficient and adaptive reconstruction process. To demonstrate the efficacy of our method, we applied it to a dataset of human faces. The results show that the method can successfully reconstruct the facial geometry, capturing details of the surfaces.