PhD Thesis Proposal Defence


"3D MODELING FROM IMAGES"

by

Mr. Ping Tan


Abstract:

In this work, we focus on the 3D reconstruction from images by photometric
stereo algorithms. We extend conventional photometric stereo methods from
two aspects. For conventional photometric stereo methods, the illumination
condition should be recorded in data capturing. Otherwise, normal
directions can only be recovered upto a General Bas-Relief (GBR)
ambiguity. In this work, we study the equivalence structure of
isotropy/reciprocity and their behavior under GBR transformation. We prove
that these symmetric structures are broken by GBR transformation. By
re-establish these symmetry structures, we can auto-matically recover the
illumination conditions. Another important limitation in conventional
photometric stereo algorithms is that surface can only be recovered at the
resolution of input images, since only one normal direction is computed
for each pixel. However, for a rough surface, there often exists subpixel
level geometry structures. We study the relationship between reflectance
and subpixel level geometry structures. We then propose a super-resolution
method for photometric stereo to compute subpixel level geometry
structures.


Date:     		Friday, 11 May 2007

Time:                   4:00p.m.-6:00p.m.

Venue:                  Room 3301
			lifts 17-18

Committee Members:      Prof. Long Quan (Supervisor)
			Dr. Chiew-Lan Tai(Chairperson)
			Dr. Huamin Qu
			Dr. Chi-Keung Tang


**** ALL are Welcome ****