Three-dimensional terrain rendering plays an important role in geography information systems, flight simulators and video games. In large outdoor environments, the visualization of terrain landscapes is challenging in terms of both storage and rendering bandwidth for the current computers. This thesis analysis the task, and describes the problems to be solved in detail.
A large-scale terrain dataset is required to implement this task, but creating one is complicated, expensive and time-consuming. This thesis show an example on how to select and obtain an appropriate dataset.
The most challenging part of the problem is the geometry management. There are dozens of publications in this topic from the past fifteen years. The thesis compares the significant algorithms, selects an appropriate one, and proposes several improvements to make the selected algorithm fit better with this task.
When working with large datasets, it is common that their size exceed the typical storage limit of memories. The thesis shows several algorithms to handle this issue, including methods that mostly use the CPU, and those that work primarily on the GPU. These algorithms are compared by their flexibility, complexity and efficiency.
The thesis also includes the implementation of the described ideas in the form of a Google Earth like application. The achieved results come close to commercial softwares in both visuals and efficiency.