Abstract:

Non-photorealistic rendering, or NPR, has emerged as an important field of computer graphics. Most NPR methods attempt to create imagery mimicking a particular style produced by an artist. Several such styles have been investigated, including painting, watercolor, engraving, pen and ink, color pencil, charcoal, cartoon coloring, stippling, and loose sketching. Thus far, most published NPR algorithms focus on a specific artistic style, or a closely related class of styles. Underlying these diverse artistic effects, however, are several recurring themes common to most NPR techniques. In this report, we present a novel framework for NPR based on adaptively sampled distance fields (ADFs). By representing a model as an ADF, we can interactively and accurately generate view-dependent particles (for stroking and coloring) and view-dependent triangles (for stroking, coloring, and visibility determination). From these view-dependent elements, many diverse styles can be realized by employing existing techniques. The multi-resolution nature of ADFs provide regulation of stroke (particle and triangle) density, guaranteed frame rates, and optimal use of processing resources in a system that scales both to new hardware and to models of increasing complexity. This approach thus unifies several operations common in artistic rendering. Furthermore, because ADFs can represent both hard surfaces and soft organic volumetric forms, opportunities exist to develop new volumetric NPR techniques in this single framework. We propose such a volumetric style.