Detail-preserving paint modeling for 3D brushes

Abstract

Recent years have witnessed significant advances in 3D brush modeling and simulation in digital paint tools. Compared with traditional 2D brushes, a 3D brush can be both more intuitive and more expressive by offering an experience closer to wielding a real, physical brush. To support popular media types such as oil and pastel, most previous 3D brush models have implemented paint smearing and mixing. This is generally accomplished by a simple repeated exchange of paint between the 3D brush and 2D canvas, with the paint picked up by the brush typically mapped directly onto the brush surface. In this paper we demonstrate that both repeated exchanges and direct mapping of paint onto brush surfaces are sub-optimal choices, leading to excessive loss of color detail and computational inefficiencies. We present new techniques to solve both problems, first by using a canvas snapshot buffer to prevent repeated paint exchange, and second by mapping brush paint onto a 2D, resolution-matched pickup map that sits underneath the brush, instead of mapping onto the 3D brush itself. Together, these act to minimize resampling artifacts, helping to preserve fine streaks and color details in strokes, while at the same time yielding improved efficiency by never sampling the brush more densely than necessary. We demonstrate the effectiveness of our method in a real-time paint system implemented on the GPU that simulates pastel and oil paint. Our method is simple and effective, and achieves a level of realism for these two media not attained by any previous work.