Level-of-Detail

Подпись: Fast Rendering of Images

The procedures discussed in the last chapter, and particularly the efficient ray­tracing implementations, allow for a relatively fast, though sufficiently exact calculation of the light distribution in complex botanical scenes. The perfor­mance of the procedures is here mostly only restricted through the quantities of the data that have to be processed.

However, it is not always necessary to show the plant models in their full com­plexity. For example, if the visual size of a plant only consists of a few pixels on screen because it is part of the background, it is pointless to use thousands of polygons for its production. The geometric resolution thus should possibly be adjusted to the visual size of the plant. If the plant is part of an animation or an interactive system and the viewer moves toward it, the complexity must change accordingly. Additionally, the modification of the complexity should not be no­ticeable. The methods used for this kind of modeling are called level-of-detail (LOD) methods. Some of these are addressed in the following.

LOD methods can be divided into two classes: static and dynamic algorithms. In a static method, for each model geometry there exists one or more reduced variants that are applied, if the model is shown in the background. Each of these variants is a unique independent form of representation with varying types of geometry. The change between individual representations occurs through over­laying, i. e., blending of the model representations. Here two representations are drawn on top of each other, where the transparency increases in one represen­tation simultaneously with the increasing distance, while the transparency in the other decreases.

In a dynamic LOD method, the representation of the model changes continu­ously depending on its projected size. The model description itself includes a procedure for coarsening or refining. Here the data is stored in such a way that precomputed parts of the model can be displayed efficiently in order to reach a reduced representation.

In practical applications, methods with a combination of static and dynamic LOD are often used. Thus, the detailed models for the foreground are mostly polygonal, meaning they are represented with their detailed surface description.

With increasing distance, these models are transformed into a dynamic LOD – model description, such as into a smaller set of surfaces or a cloud of points that can be smoothly reduced. If the model is moved to the background and is represented only by a minimal number of objects, an additional change of the representation may take place. Most often this is a flat polygon, such as a billboard, which almost always is a sufficient representation for distant objects. The classification of LOD methods depends on the underlying geometry. The surface description of a tree can be divided into two very different types of ge­ometry. The tree skeleton is a classic closed geometry with a smooth surface, which is common for representing many objects in computer graphics. In con­trast, the foliage is divided into many isolated surfaces that must be represented properly in the LOD modeling according to their character and lighting.