Techniques for efficient cloud modeling, simulation and rendering

Restricted until 1 Feb. 2010. Realistic cloud simulation and rending find applications in many 3D applications such as high-quality 3D film production and realistic 3D flight simulation. However, the infinite variation in the cloud shape and appearance make cloud simulation and rendering a challenge in termsof complexity and realisticity. The major focus of this work is on efficient and realistic cloud simulation. We adopt a similarity approach to describe the characteristics of simple fluid flow without rigorously solving the governing PDEs. Based on the similarity solution of turbulent thermals, plumes and vortex rings in stable stratified fluids or unstable environments, we propose three methods for cloud simulation and modeling.; First, a two-stage cloud simulation method is proposed to model the cloud formation process. At the first stage, instead of solving PDEs for the velocity profile, cloud formation is obtained using a lower level building block; namely, particle simulation. New attributes of particles are incorporated to meet the modeling requirement, {\em e.g.,} the condensation time and density of particles. For particle motion, the characteristics of the convective process is captured by the thermal similarity solution consistent with the governing Navier-Stoke equation. At the second stage, we consider the process of being condensed into the visible water droplet. The voxel is used to store the water vapor density and the water droplet density. Due to the release of latent heat, we choose to have a small scale convective process. Finally, the cloud droplet density is rendered using a two-pass rendering method and multiple scattering to result in the high albedo property of the cloud.; Then, a decoupling cloud simulation method that divides three spatial dimension into independent horizontal and vertical dimensions is proposed. With the assumption that the radial profiles at all heights are similar, we choose several key frame layers only for actual 2D calculation, interpolate other layers from them and add some turbulence. Along the horizontal dimension, the movement is driven by heat difference. This virtual temperature difference results in a different water vapor distribution. To model the cloud formation from a plume, we assume that the water phase change occurs at its late development stage, where the plume spreading between levels of buoyancy vanishes and the vertical velocity is equal to zero. At such a stage, the water vapor will be sufficiently accumulated to experience the water phase transition. The 2D cloud layer is then simulated by the Navier-Stoke equations. Besides the momentum equation coming from the virtual temperature difference, one more force term is introduced in our model, i.e. the Coriolis force. Finally, water droplets are rendered using a 3D texture rendering technique. By separating the horizontal and vertical couplings in the traditional 3D simulation, we can reduce the computational cost yet capture the essential characteristicsof the cumulus cloud.; Finally, we introduce a novel hierarchical cloud simulation method. The general shape and movement of clouds in the simulation model are governed by the similarity of thermals. Furthermore, their relative position to the camera is calculated. If they are not hidden by other thermals or they are in the boundary of a cloud, we turn on thermal's inner motion which behave as a weak vortex ring. We simulate thermal's interior dynamics using the axis-symmetric flow, which provides a simple scheme to determine the velocity at any arbitrary location inside the flow. With the axis-symmetric flow, we adopt the Lagrangian approach for thermal's interior motion, where a bunch of water droplet particles inside the thermal are updated at each time-step, generating the desired cloud detail structure. Furthermore, we consider the phase transition with a simplified strategy. With the entrainment mass whose coefficients are chosen by users, we can increase the water droplet's density, thus its size and transparency for rendering. A method for cloud formation due to the latent heat release from the condensation is also considered.