As a part of a driver assistance system, vehicle top-view system plays an important role in driving safety. The objective of this research is to develop a vision-based top-view system to expand drivers' field of view. Four cameras provide a 360-degree view of the surrounding terrain of the vehicle, and the destination image is composed by these four processed images. In this paper, we divide the procedure into 4 parts, fisheye lens distortion correction, inverse perspective transformation, image mosaicking and pixels mapping table creation. Every part is achieved by one-to-one mapping between pixels, so most of the calculation has been done on PC. The experimental results show that, the final image is clear, natural and have very good real-time performance. It can effectively reduce the blind spots of drivers, and it has a strong value for research and practical.