Real-time Rendering Of Large-scale Water Scene

With the popularization of virtual realities, movies, animations and video games, faithful simulation of large-scale water scene has become a hot and challenging topic in computer graphics community. Although there is much work on rendering water scenes, it is still challenging to simulate interactive water scene due to its complexity of modeling and rendering.In this thesis, we introduce a GPU-accelerated algorithm for rendering large-scale water scene. We apply an FFT method, which is based on a statistical model, to the simulation of ocean wave. Through this method, we not only compute 3d displacement data for surface grid of water, but also get the slope data, which can be used in rendering of water. We apply "GPU-based wave equation simulation" and "particle-system-based simulation" to the simulation of ship wake, and we also show their pro and cons.We employ an LOD-friendly "projected grid" technique for the grid generation. To make this grid generation algorithm amenable, we further introduce a "projected texture" to improve the rendering quality. We also improve "projected-grid camera" to make it adaptive to non-horizontal camera and height-variant water scene as well as to reduce the alias effect of displacement textures.In the rendering part, we unify the rendering of both ocean wave and ship wake by joining their height field data and slope field data. We fully consider the optic characteristic and HDR effect of water when doing the rendering. Foam, spray and caustics further enhance the rendering quality.To make our water simulation and rendering real-time, grid generation, height field calculation, slope field calculation, grid displacement and rendering are all processed by GPU, which greatly reduce the synchronization of CPU and GPU.In Chapter 1, we give a brief review of the related work. We then introduce simulation algorithms for ocean wave and ship wake in Chapter 2. In Chapter 3, we present our method for generating grid and height field, and discuss the scheme to improve "projected-grid camera". We describe our rendering approach in Chapter 4. The integrated simulation pipeline and experimental results are given in Chapter 5.