Surface Estimation from Multiple Images

Mads Kessel

AbstractThis thesis presents an overview of the general problem of surface estimation from 2D images. It describes the currently known approaches, and states the pros and cons of them.

An algorithm proposed by Vogiatzis, using deformable models in a Bayes framework for estimating the surface from a set of images with known intrinsic and extrinsic parameters, is thoroughly described and reflected upon. The algorithm is implemented in C++ using OpenGL and the OpenGL Shading Language to help performance. A number of deviations to the work of Vogiatzis is proposed and implemented as well.

The implementation is thoroughly tested in both a synthetic environment, where the ground truth is known, and using several real world datasets. The result is used to conclude on the algorithm by Vogiatzis, and the proposed deviations to it.

Several conceptually new proposals are described, some of which has been implemented and evaluated.
KeywordsSurface estimation, 3D Reconstruction, Structure from Motion, Deformable Models, Simulated Annealing, Bayes analysis, OpenGL, GPGPU, Image Metric, Surface constraints, Mesh Simplification
TypeMaster's thesis [Academic thesis]
Year2006
PublisherInformatics and Mathematical Modelling, Technical University of Denmark, DTU
AddressRichard Petersens Plads, Building 321, DK-2800 Kgs. Lyngby
SeriesIMM-Thesis-2006-43
Note
Electronic version(s)[pdf]
BibTeX data [bibtex]
IMM Group(s)Image Analysis & Computer Graphics