Non-Photorealistic Computer Graphics Library

Integration of Non-Photorealistic Rendering Techniques for 3D Models in Processing

Author(s): Katrin Lang.
Master Thesis: Technical University of Berlin, May, 2009.
[BibTeX]

Abstract:
This thesis presents the implementation of a non-photorealistic renderer for the pro- cessing API (Application Programming Interface). For decades, the goal of traditional computer graphics has been to create artificial images of simulated 3D environments that look precisely like a high-quality photograph of the real world. Yet, this may result in an ``overkill'' of visual information, when in many cases abstraction is the more appropriate way to convey shape and meaning. Examples include technical, scientific, medical and narrative illustration, computer aided design, architecture, and computational art. In contrast to photorealistic rendering, the branch of computer graphics called NPR (Non-Photorealistic Rendering) is inspired by styles derived from painting, drawing, and animated cartoons. Several of these techniques have made their appearance in video games and movies, most prominent among them cel or ``toon'' shading, and have been integrated into professional commercial rendering packages. However, the state of the art in NPR exists only in research papers and experimental academic implementations. There are virtually no freely available applications that practicioners could use to experiment with such techniques. Conversely, Processing lacks the ability to create meaningful abstractions in 3D. Although it allows for a wide range of expressive styles for digital art in 2D, 3D techniques are for the most part limited to Gouraud shading. Being an open source platform with a wide community and offering a comprehensive and easy-to-use graphics API, Processing is the ideal platform for making latest developments by researchers in non-photorealistic rendering available to designers and artists. The Processing NPR renderer presented in this thesis is completely hardware based, as opposed to the original Processing renderer, but the interface remains the unaltered Processing language. It comes with a comprehensive set of NPR algorithms for shading, shadows and line drawings, which are carefully designed for interactive applications. Moreover, this thesis provides a framework of NPR data structures and rendering infrastructure that allows NPR researchers to plug in their algorithms in order to gain valuable feedback by users and create a synergy effect between researchers and practicioners.