In computer graphics – especially in computer games – one of the most important and fundamental elements of the virtual scene is the terrain. Virtual terrains vary widely depending on the needs of the application: a real-time stategy game puts different constraints on the virtual terrain model, than for example a flight simulator. In one application, the size and distribution of flat regions (to build on) are important, while the other needs large, realistic landscapes. Although their use-cases differ, these terrain models share a common element: they are never crafted by hand, but generated with an appropriate software tool.
To satisfy the ranging needs, many different types of terrain generators have been developed over the time. Most of these generators use a small set of input parameters, that mosty affect the terrain globally and are counter-intuitive to use.
This problem could be solved by a special kind of terrain generators: a concept-driven one. By using this kind of generator, the user defines a rough sketch, the „concept”, of the terrain, by a small number of intuitive, visualizable primitives: hills, cliffs, mountain ranges, riverbeds, lakesides. The generator then takes this concept and tries to create a landscape that fits these constraints.
In this thesis we review the most popular and fundamental algorithms and concepts that are used in modern terrain generator applications. We evaluate their strengths and weaknesses, then we present a terrain generator application that is capable of creating landscapes, based upon the user’s initial concept.