Abstract:
We present a novel approach for adaptively grouping and subdividing hair using discrete level-of-detail (LOD) representations. The set of discrete LODs include hair strands, clusters and strips. Their dynamic behavior is controlled by a base skeleton. The base skeletons are subdivided and grouped to form clustering hierarchies using a quad-tree data structure during the precomputation. At run time, our algorithm traverses the hierarchy to create continuous LODs on the fly and chooses both the appropriate discrete and continuous hair LOD representations based on the motion, the visibility, and the viewing distance of the hair from the viewer. Our collision detection for hair represented by the proposed LODs relies on a family of "swept sphere volumes" for fast and accurate intersection computations. We also use an implicit integration method to achieve simulation stability while allowing us to take large time steps. Together, these approaches for hair simulation and collision detection offer the flexibility to balance between the overall performance and visual quality of the animated hair. Furthermore, our approach is capable of modeling various styles, lengths, and motion of hair.
