|
|
 |
|
www.balmelli.net
|
Rate-Distortion Optimal
Mesh Simplification for Communications
Laboratory
for Audio-Visual Communications
Ecole
Polytechnique Federale, Switzerland
Phd dissertation No
2260, 2001
Advisor:
Professor
Martin
Vetterli Co-Advisor: Professor Thomas Liebling
download
pdf version
Topics (~chapters):
(1)
A
transmission
system for triangular mesh.
We describe a conceptual
transmission system for texutred meshes and identified the
problem
to address in order to obtain an progressive and
bandwidth-adaptive
transmission.
(2) A
journey in piecewise-linear function approximation.
We review a set
of optimal and greedy simplification strategies. We use the simple
polyline
model and evaluate the computational complexity of each approach. Then,
we compare each method is
term of approximation quality across rate (measured as the number of
polyline
knots).
(3)
Computational analysis of 4-8 meshes.
We
introduce a type semi-regular mesh whose subdivision connectivity is
obtained
with the 4-8 scheme. We analyse a set of simplification operations:
vertex
insertion and decimation. Then, we explain several properties of the
mesh
construction. These propertis are useful for geometry simplification
algorithms.
(4)
Quadtree data structure for efficient storage and access of 4-8
meshes
We
introduce a quadtree data structure to store efficient semi-regular 4-8
meshes. To quadtree store the 4-8 mesh without redundancy and provide
an
efficient constant-time access mechanism to query and process the
dataset.
(5)
Progressive meshes in an operational rate-distortion sense.
Given
a semi-regular mesh whose subdivision connectivity is obtained with the
4-8 scheme, we use a tree-driven, fine to coarse approach to simplify
the
mesh using vertex decimation. Our method uses global error and a
generalized
vertex decimation technique borrowed from optimal tree pruning
algorithms
used in compression. The output representation is adaptive and
progressive.
(6)
Joint Mesh-Texture optimization.
Given
a bit budget to spend on the representation of a textured polygonal
mesh,
how can the texture and the geometry information be balanced in order
to
maximize the rendered quality? We propose a greedy, yet efficient
optimization
method to address this problem.
|
|