M.Sc. Thesis Defence: A Comparative Study of Spectral Embedding Methods - Xiaoming Li
Spectral methods, which employ eigenvalues, eigenvectors, or eigenspace projections derived from linear operators, have been proposed in the computer science literature in recent decades. In the area of geometry processing and analysis, various spectral methods have been developed and used to solve a diversity of problems, such as shape classification, graph partitioning, mesh parameterization, mesh segmentation, shape correspondence, and symmetry detection.
In order to have a better understanding and use of the strength and weakness of different spectral approaches, this preliminary comparative study will explore the behaviors of four spectral embedding methods: Global Point Signatures Embedding (GPSE), Heat Kernel Signature Embedding (HKSE), Multi-Dimensional Scaling Embedding (MDSE), and Spectral Embedding using Gaussian-filtered affinity matrices (SEG), working on three applications: segmentation, correspondence and symmetry detection. The goal will be to observe and investigate the similarities and differences of spectral methods when applied onto different applications.
Keywords: spectral embedding, geometry processing, 3D shape, mesh segmentation, symmetry detection, shape correspondence
M.Sc. Examining Committee:
Dr. Hao (Richard) Zhang, Senior Supervisor
Dr. Torsten MÃ¶ller, Supervisor
Dr. Ghassan Hamarneh Examiner
Dr. Ramesh Krishnamurti , Chair
TASC1 9204 West
Tuesday, October 30, 2012 - 10:30