The main goal of this work is to address the aforementioned gaps in the computer-aided landscape visualization by systematically bringing the determinant features of traditional media at the level of modern, state-of-the art computer graphics. The ultimative goal is to create interactive computer visualizations of vegetation, complex landscapes as well as environmental data that comply with the requirements of the landscape practice. There have been precise guidelines and criteria to follow during the work:
• widely available GIS(Geographical Information Systems) data should be used to create visualization of vegetation, complex landscapes as well as environ-mental data automatically. As GIS and additional compatible software (the Lenne3D visualization system, www.lenne3d.com) are available, this would ensure easy integration in practice.
• 3D computer generated imagery should convey an easily and quickly under-standable impression of a landscape from the most usual perspective to the normal audience - walking perspective, at a suitable, user-defined, detail level.
1.2. GOALS 5
• in contrast to conventional renditions, it should be possible to control the ab-straction degree, such as to abstract out unimportant details that only detract the observer and focus the view on the essential landscape characteristics that have to be transmitted.
• the algorithms should be as real-time capable as possible and should use the potential of modern PC graphics cards on consumer PCs (without requiring very expensive or dedicated hardware), so that to allow interactive rendering of complex plant and landscape models.
• user-controlled separation and differentiation of different landscape elements should be possible, in order to allow meaningful visualizations that convey certain information (such as the current and planned state).
• intuitive, in-situ visualization of environment-related data and landscape char-acteristics should be made possible. This would be a great benefit in commu-nication, since such data is not "visible" in a pure landscape visualization or it is beyond common understanding.
The approach of the current work builds on non-photorealistic rendering. A flexi-ble landscape visualization framework according to the aforementioned criteria is to be developed, using both existing techniques, improvements as well as new ideas.
However, the transposition of traditional landscape media principles into computer graphics techniques presents serious technical challenges that have to be addressed:
• existing NPR algorithms are usually designed for specific applications, objects and rendering styles. Applying them to landscapes may require substantial adaptation and improvement. Moreover, plants constitute a particular class of objects: due to their fragmented nature, they are topologically very differ-ent to the more common three-dimensional compact objects, like buildings for instance. This may cause difficulties for existing algorithms aa additional re-quirements must be fulfilled.
• mostly, the computational complexity of NPR algorithms, even of those that run at interactive speeds, is significantly higher than hardware-accelerated con-ventional real-time rendering. Thus, they can mostly handle only small and simple scenes. Because interactivity is an important criterium in the current application, a proper balance between quality and efficiency must be achieved.
• except for simple cases, NPR algorithms are not hardware-accelerated (they do not take advantage of the significant processing resources of modern graphics hardware). This, however, is an important issue for achieving sufficient speed for real-time display of large landscape scenes.
• Not only single plant models should be handled, but also especially large plant groups (forest, meadow), for which there are currently no satisfying ap-proaches(see Fig. ).
• it is not straightforward to combine different NPR algorithms, as they can be built using very different principles. Allowing style combination is, however, an useful feature, as it substantially extends the expression possibilities in com-munication processes.
• extensive user control on the rendering appearance is required if landscape visualization should be used as a communication tool, while many NPR algo-rithms offer pre-determined styles. A mapping of user interaction to parame-terized NPR algorithms must be developed.
Figure 1.3:Top row: computer generated photorealistic(left) and sketchy(right) landscape images. Bottom row: sketchy landscape structures (left) and environment data visualization (right). All images are snapshots from the Lenne3D Player soft-ware using the framework described in the current work.)
Although, in its substance, it is mainly focused on computer graphics visualization techniques, the current research has a strong inter-disciplinary character and has been carried in an inter-disciplinary setup. All techniques have been developed with the final purpose of improving and extending visualization media in disciplines like land-scape planing, architecture as well as their applications related to environment. From another point of view, experts and expertise in these various fields have decisively contributed to the realization of this thesis.
1.3 Specific Contributions
The main contribution of the current research is a set of customizable, real-time non-photorealistic rendering algorithms (three-dimensional interactive sketches) that attempt to provide a solution to the meaningful landscape visualization in the context of landscape planning, architecture and environment communication and education.
It is the result of considering recent advances in computer graphics technology, as well as the conclusions of different surveys among the potential users, namely land-scape and environment experts.
The core of our framework are computer-simulated pen-and-ink illustration tech-niques, which we call interactive sketches. The choice of sketch, and particularly of
1.3. CONTRIBUTIONS 7