The Hong Kong University of Science and Technology
Department of Computer Science


PhD Thesis Defence


"3D Structure Computation from Multiple Views"

By

Mr. Yichen Wei


Abstract

3D structure recovery from images is fundamental and extensively studied
in computer vision. This thesis investigates this problem from multiple
images taken at different viewpoints about a rigid 3D scene. We tackle and
contribute to several problems in different stages. In the preliminary
stage of camera calibration, we work on camera motion estimation under
circular motion. An effective algorithm is developed and suitable for a
large image sequence.

In the 3D reconstruction stage, dense representation of the scene is
computed from calibrated images. We firstly propose a three-aspect
categorization that differentiates previous 3D reconstruction methods
based on their various properties. Then we study two different problems.
In the first problem, depth maps are used to represent the scene and
computed from multi-view. A novel formulation with asymmetrical occlusion
treatment is proposed and two approximate optimization methods are
developed. Our approach produces comparable results and is more memory and
time efficient compared with previous methods, therefore suitable for
multiple input images. The second problem is image-based hair modelling
arising from needs in computer graphics applications. It differs a lot
from traditional problems due to the special property of hair geometry. We
developed a practical system based on multi-view. It is highly automatic,
convenient to use and produces realistic models of real hair.


Date:			Wednesday, 23 August 2006

Time:			2:00p.m.-4:00p.m.

Venue:			Room 5501
			Lifts 25-26

Chairman:		Prof. Kun Xu (MATH)

Committee Members:	Prof. Long Quan (Supervisor)
			Prof. Chiew-Lan Tai
			Prof. Chi Keung Tang
			Prof. Rama Chellappa (Comp. Sci.,
					      Univ. of Maryland)
			Prof. Richard Hartley (Information Engg.,
					       Australian National Univ.)



**** ALL are Welcome ****