July 22, 2013

Simulations in Tinseltown

Computer graphic scientists at IST Austria present new simulations at the most important scientific conference in their field

In this liquid simulation, commonly used models for animation dissolve in splashes and waves. (© Ryoichi Ando)
In this liquid simulation, commonly used models for animation dissolve in splashes and waves. © Ryoichi Ando

Disneyland seems an unlikely venue for a scientific conference but the home of Mickey, Donald and Captain Jack Sparrow is where IST Austria professor Chris Wojtan is going to present three papers by his group and other co-researchers.  Anaheim, California, is not only the home of the first Disneyland, but also annually hosts the leading computer graphics conference SIGGRAPH, taking place this year on July 21-25.  The dramatic crushing waves of the “Pirates of the Caribbean” by Disney Studios are thanks to computer animations, which scientists are striving to make ever more convincingly realistic. But this is only a side aspect. The main objective of the computer scientists at IST Austria is to design deceptively real simulations which are essential for using animations in other fields of science, like animating moving cells or embryonic development.

Morten Bojsen-Hansen, PhD student in the group of Chris Wojtan, presents the paper “Liquid Surface Tracking with Error Compensation” together with Chris Wojtan (see http://pub.ist.ac.at/group_wojtan/projects/2013_Bojsen-Hansen_LSwEC/index.html for demo video). Their paper addresses the problem of accurately tracking and simulating the surface in fluid simulations, such as water droplets arising from splashing water and waves. To track liquid surfaces, simulators use a computational machinery called “surface tracker” on top of the liquid simulation simulating the underlying bulk of fluid. To make a detailed and visually rich animation, details need to be added to the surface tracker. One option is increasing the resolution of the surface tracker responsible for the surface animation, but keeping the underlying liquid simulation at low resolution to reduce the computation power required. However, this previously used solution adds visual and topological artifacts that are not realistic, making it unusable for highly detailed animations. In their paper, the researchers present a new solution allowing for visually rich and highly detailed fluid simulations that are physically accurate. They find the theoretical basis for visual artifacts that arose with previously used methods, and present a novel way for quantifying whether a fluid surface tracker is physically valid. They then develop a novel algorithm that removes the artifacts, while preserving valid details.

Ryoichi Ando, Visiting Student with the Wojtan group and PhD student at the Graduate School of Design at Kyushu University in Japan, together with Nils Thürey at Scanline VFX and Chris Wojtan presents a paper entitled “Highly Adaptive Liquid Simulations on Tetrahedral Meshes” (http://pub.ist.ac.at/group_wojtan/projects/2013_Ando_HALSoTM/index.html). In their work, the researchers introduce a new method for simulating liquid with extreme amounts of spatial adaptivity. Their aim was to enable simulators to efficiently create high-resolutions animations while decreasing the amount of computation required. Their new simulation algorithm allows simulators to focus computational resources on those regions of the animation that are visually interesting, while avoiding artificial and unnatural distortions in the flow of fluid. In their demo video, they show examples of how their simulation algorithm can animate very large surfaces of liquids with highly localized details, such as highly detailed wave splashes.

Finally, Chris Wojtan and Gilbert Louis Bernstein of the University of Washington and Stanford University present a method for computing topology changes in interactive modeling environments (“Putting Holes in Holey Geometry: Topology Changes for Arbitrary Surfaces”; http://www.gilbertbernstein.com/project_toptop.html). Programs for modeling surfaces need to support ways to change the topology, for example to put a hole into a sphere to model a donut. The authors developed a method that can be used to change topology, which also works in models that do not have the geometric properties required by other programs to change topology.  Incorporating a novel scheme for correcting errors, the newly developed method is able to tolerate a variety of mistakes in the surface. As the new method is easily integrated into geometric modeling applications, expectations are that it will significantly improve the work flow in geometric modeling.


Back to Top