Three-dimensional visualization is one of the most important researches in the field of medical images processing. Volume rendering is a powerful technique for visualizing meaningful information extracted from volumetric data in three-dimensional visualization. Although it does not need to generate intermediate geometric primitives and its rendering result is high quality, computational complexity of volume rendering is too high to achieve real-time interaction. Therefore, a fast GPU based high-quality three-dimensional visualization volume rendering method is proposed. A GPU based method is proposed to improve the run-time efficiency of volume rendering using volumetric texture. Because of limitation of video memory storage, large-scale volumetric data cannot be reconstructed or an expensive GPU is required. Therefore a slice texture based volume rendering method is proposed, which merely cost 4MB video memory by making use of a real-time dynamic switch mechanism. On the other hand, in order to eliminate annular distortion of volume rendering, a GPU based pre-integration classification method is presented. Although our methods are all implemented at general PCs, they can not only guarantee image quality but also improve rendering speed.