Overall Objectives
Scientific Foundations
Application Domains
New Results
Contracts and Grants with Industry
Other Grants and Activities
Inria / Raweb 2003
Project: EVASION

Project : evasion

Section: New Results

Multi-Resolution Animation

Participants : Florence Bertails, Marie-Paule Cani, Fabrice Neyret.

Advected textures

Participant : Fabrice Neyret.

In order to add animated high-resolution details to a coarse fluid simulation, we developed a method which consist of advecting a flow noise texture (a previous technique proposed by Perlin and Neyret) in such a way to avoid the classical flaws such as stretching and ghosting artefact (see figure 10). The vorticity of the flow noise procedural texture is controlled by relying on a Kolmogorov cascade model conveying the energy from the highest spatial frequencies of the coarse simulation. Various parameters let the users control the behaviour of the animated texture. This work has been published at the Symposium on Computer Animation'03 [16].

Figure 10. Coarse simulation and its dressed counter-part. Three 2D and one 3D examples (note that the corresponding 3D grid is only 8 × 8 × 8 ).

Multi-Resolution Particle Systems

Participant : Fabrice Neyret.

We have worked on the phenomenological simulation of avalanches, in the scope of the "slides and avalanche" local-ARC collaboration (see section  8.2). The model consists of hierarchical particles implementing the various passive and active elements such as the front vortex, point and tube vortices existing at various scales, and fluid markers. Main particles are only affected by the coarse fluid simulation (or by an autonomous dynamics model) while lower particles are affected by the velocity field generated by the higher level vortex particles (see figure 11). We are currently working on coupling the Advected Texture method to this by associating texture coordinates to the marker particles (at the lowest level).

Figure 11. The visualisation of passive marker particles reveals the action of vortex particles of various scales acting on them.
avalanche1 avalanche2

Adaptive hair animation

Participants : Florence Bertails, Marie-Paule Cani.

The layered models we had experimented in the past for hair animation were far from real-time, due to the number of interacting hair wisps needed to capture the complexity of hair motion (see figure 12). To gain in efficiency, we recently proposed an adaptive animation technique, according to the methodological principles listed in section 3: hair wisps subdivision (where complex motion is observed) and fusion (where wisps have the same motion than their neighbours) enable the hair to automatically refine when and where needed [11]. This method results in interactive rates, and has been selected for an industrial transfer of technology (see sections 7.3 and 7.4).

Figure 12. Comparison between real hair and our adaptive hair simulation.