"Speed the rendering of 3D terrain in computer games"

Supervised by S.Maddock

Abstract

Computer games today feature complex terrain as a main part of the game environment. It is important that the terrain is realistic and is varied. A good terrain will help prolong the longevity of a game, and it is therefore important to make the terrain as close to the "real thing" as possible to gain and maintain the interest of a game player.

Rendering a large terrain takes time. There are at least 10s of thousands polygons to render. In real-time applications speed of rendering is a very important factor. Real-time means that several frames are being rendered per second. Even with help from specialised hardware the amount of polygons that make up a good terrain, are too large to be considered every time the frame is rendered. There exist several algorithms that will reduce the set of polygons and preserve the quality of the terrain, e.g. level of detail (LOD) and potentially visible sets (PVS).

This project set out to find a new possible algorithm that will reduce the number of polygons that are rendered on each pass. The project will investigate whether pre-calculated information is possible solution to a fast algorithm that determines which parts of the terrain are visible. The application provided a good test suite so that all the necessary information about the algorithm's performance could be extracted. It also provided another algorithm for comparison. The results were compiled and investigated and concludes that the algorithm does not produce a satisfactory result.