Non-Photorealistic Rendering Techniques Based On Programmable Graphics Hardware

Non-Photorealistic Rendering (NPR) is a newly developed research area in computer graphics and, in some cases, NPR is more attractive than photorealistic rendering. Thus, it has a wide range of applications in the fields ranging form education, entertainment to art. NPR aims at generating stylized images in 2-D space or 3-D space and can be classified into image based or model based techniques.In the early stage of NPR, the algorithms are implemented using software and executed in CPU. In order to take advantage of hardware power, some researchers began to use display card to accelerate the NPR execution. In traditional graphics hardware, however, 3-D vertex data are processed through a fixed-function graphics pipeline. This doesn't meet the requirements of NPR, because NPR requires flexible control when generating different styles. The emergence of a new technology called programmable graphics pipeline sheds the light to this problem. The flexibility of the programmable graphics pipeline is achieved by adding new programmable modules into traditional graphics pipeline. Based on programmable graphics pipeline, display cards are able to accelerate graphics rendering while still keeping flexibility. .The research in this thesis focuses on Non-Photorealistic Rending techniques based on programmable graphics hardware. We introduced and implemented a NPR system for mountains in Chinese landscape painting style. With this system, we can generate a variety of image series in real time.