Modélisation interactive par points d'objets complexes à partir d'images
François Duranleau
M.Sc. Thesis, August 2002
Abstract
Many techniques have been developed to interactively display highly complex
graphical objects (forests, mountains, cities, etc.) or simpler objects but with
high precision (people, sculptures, vases, etc.) by using 3D scanners. One of
these display techniques uses points to sample the objects themselves. Based on
this idea, we present an automatic and interactive system which allows to
reconstruct complex geometric models. We rely on a 3D point-based stochastic
reconstruction method which uses a limited number of images without constraints
on their pose. Our results show that a point-based representation and the
stochastic aspect of our reconstruction method are efficient to grasp the
complexity of real objects' details.
Another important aspect of our system is its interactivity. The complexity of
objects that we attempt to model induces great difficulties for most 3D
reconstruction methods based on feature correspondance in images. In our system,
it is possible for the user to guide the reconstruction process. We have
observed that it allows to speed up and to improve the final result.
Online version
Available here in french (en français)
in Adobe PDF format (31 MB).
BibTeX
@MastersThesis{Duranleau:2002:MSC,
author = "François Duranleau",
title = "Mod{\'e}lisation interactive par points d'objets complexes {\`a} partir d'images",
month = aug,
year = 2002,
type = "M.Sc. Thesis",
school = "D{\'e}partement d'Informatique et Recherche
Op{\'e}rationnelle, Universit{\'e} de Montr{\'e}al",
}