Depending upon the area of application, different goals are pursued in landscape visualizations. In botany often only an approximate representation of the model is required. In architecture, and in some cases also in landscape planning, in computer games, and in simulation, images are required that, in addition to form, reflect light conditions as well as very detailed structures. The film and advertisement industry demands photorealistic images for merging real-life characters with synthetic scenes, and for unique landscaping. Another abstract type of representation is required for cartoons, where images in comic style are needed.
Thus the question arises: what effects are necessary for generating synthetic landscapes? For example, is a complex simulation of the global illumination necessary for the computation of photorealistic images of plants? How do the different illumination factors affect the related “authenticity” or “reality” of the representation?
If approximate geometry representations are to be used for the reduction of the data, more questions emerge: What actually gives vegetation its characteristic appearance? In what places are we able to replace geometric detail with prefabricated pictures or textures? And how can the “area-fill-mass” in a large scene be represented? Furthermore, how do level-of-detail descriptions, meaning those that render a plant differently in detail depending on its projected size on the screen, integrate into the process, i. e., the rendering? What kinds of methods allow us to reduce the geometric quantity, meaning the quantity of the entire data that has to be produced and analyzed and processed? Moreover, how can the transition between different model representations be implemented without the viewer becoming aware of it?
Finally, it is useful to clarify what actually constitutes a “beautiful” landscape. This is an intensively researched area in the field of ecology as well as in the area of aesthetics. What can be derived from the expression “beautiful” in order to create a corresponding visually beautiful landscape and how can already created landscapes be improved? Are there any global limits for the number of different plant models needed in a scene?
We already mentioned that even the nonphotorealistic representation of a synthetic landscape has a distinct function. Accordingly, in landscaping and architecture, the results of planning are traditionally represented as sketches. Some of the reasons are that a sketch visualizes a model of an object, whereas a photorealistic representation renders a concrete instance. Additionally, it is easier to convince viewers to make changes in a sketch rather than in a realistic representation.
Also, the sketch does not confine the planner to a certain appearance, as would a realistic-appearing computer image. Yet, with the sketch, size and light conditions can be rendered more accurately. Furthermore, sketching makes it possible to emphasize important areas using contrasting methods, and thus to direct the focal point of the observer.
Chapter 1 On the contrary, the abstract visualization of plants requires new methods for
Computer-Generated Plants the production of images. Thus, one needs specially prepared models, in which
the geometric data is reduced and changed with regard to the representation. There are a number of questions, some of which also have been of concern to the art community: how is form, and how are light and shadow represented? In Chap. 11 we show that there are numerous different rendering methods available for plants in traditional arts. While some of these techniques are applicable to synthetic rendering, for the most part completely new approaches are necessary, and structures of required geometrical models have to be changed accordingly.
Abstract representations increase the potential for visual expression and communication in computer graphics a great deal. In the future, the user will be able to choose the most appropriate rendering method from a wide variety of possible representations. The photorealistic representation is here only one of many options.