Artists and illustrators have developed a large repertoire of techniques to communicate information effectively using traditional media. Recent work in computer graphics has begun to leverage these techniques in the form of non-photorealistic rendering (NPR) algorithms for 3D, but, to date, little research has addressed flexible, interactive tools to make such algorithms accessible to content creators. In this dissertation we demonstrate the importance of developing these tools for NPR. In particular, we show that “hands-on” NPR systems can provide the designer with new levels of æsthetic flexibility, and the means to achieve effects tedious or even impossible to attain by traditional methods.

We approach this open problem for stroke-based NPR of animated 3D geometry. Our system employs a tablet interface to provide the designer with an interactive paradigm in which stroke effects are sketched directly into the scene. The artist imparts his unique æsthetic by sketching strokes over the outlines of objects, and drawing details and hatching effects onto their surfaces.

These tools raise new questions regarding how annotations drawn by the designer in one view should appear when the scene is rendered from novel viewpoints. It is unreasonable to expect the user to annotate geometry from all conceivable poses. Rather, we develop rendering algorithms that can exhibit a range of behaviors which the designer can tune to specific goals, either by adjusting parameters or directly sketching the desired effects. But it is not sufficient for the stylization rendered in each frame to maintain consistency only with the designer’s annotations. To be suitable for animation, they must also exhibit coherence between consecutive frames. We consider this for all effects in our system, but this is particularly challenging for those stylizing view-dependent silhouettes. For these, we develop a flexible and robust new framework that can provide coherence tailored to a particular æsthetic. These tools make possible a new class of effects that would be daunting to animate by hand.

This research shares applications with those of traditional hand-generated techniques, including architectural and product design, technical and medical illustration, storytelling (e.g. children’s books), games, fine arts, and animation.