Abstract
The complexity of model influences the quality of rendering. However, even the latest powerful graphics hardware can not afford too many models with high polygon count in the scene. In computer graphics, a depth map which encodes the information of height field surface may be used to displace original geometry surface. Therefore, the rendering of low polygonal surface can reach the same quality as a more complex model of much higher polygon count. There are several effects caused by height field surfaces. Among these effects, the silhouette is an important feature but is often neglected in the rendering. In this thesis, I proposed an image based technique which uses geometry normal and tangent maps to render surface details such as motion parallax, self-shadow and silhouette at interactive rates. I also provide in-depth survey of related methods and summarize their respective capabilities and drawbacks for comparison.