An Adaptive Particle Level Set Method

Fluid animation is one of the most desirable techniques in computer graphics and virtual reality. As these phenomena contain highly complex behaviors and rich visual details, it is difficult to deal with the complex motion of the water-air interfaces. Therefore, capturing the interface accurately and efficiently is a key issue in fluid animation. In order to address the problem of high numerical diffusion and low efficiency in traditional methods such as level set and particle level set algorithms, an adaptive particle level set method is presented. The particle placement in our approach is modeled as a stochastic process. Desirable goals are then achieved by allocating more computational resource to regions of high numerical dissipation during animation heuristically. In order to derive the optimal-rules of computing the stochastic process, we construct an importance sampling model and evaluate the volume loss in computational domain. The probability density function (PDF) of particle placement is obtained by employing a geometrical-based sampling algorithm which adopts a novel local feature size function on narrow band level set. The efficiency of computing this stochastic process is further improved according to the definition of accumulated shape deformation. Experiments show that the proposed approach provides high quality and low cost both in numerical tests and water animation applications.