Mesh Processing and Optimization for Scientific Visualization
Markus Gross (ETH Zürich)
In recent years powerful low-cost computing resources continuously push the size
and complexity of scientific and engineering simulations. As a consequence,
contemporary scientific visualization methods have to cope with graphics objects
of tens of millions of triangle primitives. In spite of the advances made in 3D
graphics hardware these highly detailed models still pose a great challenge for
interactive rendering. As conventional algorithms approach their limits, more and
more sophisticated mesh processing techniques are developed to improve the
quality and performance of visualization methods.
In this talk I will give an overview of mesh processing and optimization
algorithms as they are used in the graphics and visualization communities for
surface and volume representation. I will start with the concept of progressive
meshes that computes a multiresolution sequence of triangle meshes by using a
topological operator called edge contraction. The method works for surface and
volume meshes. The coarsification is governed by an energy function that allows
us to control individual data features including gradient, volume or shape.
In many applications raw input meshes have to be smoothed prior to visualization.
An often used strategy includes so-called recursive subdivision providing a
hierarchy of refined meshes that converge asymptotically to a smooth surface
representation. As such subdivision combines the advantages of piecewise linear
functions with higher order polynomial surfaces. Alternatively, fairing methods
smooth meshes by iteratively solving a diffusion equation on the surface. In the
second part of the talk I will discuss the smoothing of manifold and
meshes, feature detection and preservation.