Real-time image-based stylistic rendering using graphics hardware acceleration

Abstract

Silhouette is an important research issue in the field of Non-Photorealistic Rendering (NPR) and it is also a popular drawing feature in illustrations and line-drawing artworks. In this paper, we present a real-time image-based stylized rendering system. First, we project a 3D model to image-space. Then we extract edges in the image-space data. We perform edge-detection algorithms on GPU (Graphics Processing Units) for speedup. GPU is good at floating-points calculating and processing with parallelism. Both features match the property of most image processing tasks. Our system can run at an interactive frame rate when combining our edge-detection algorithms with graphic hardware architecture. We demonstrate that this system performance can reach real-time and render images in good NPR style. © 2007 IEEE.