Rendering realistic human hair is avery challenging and important issue in computer graphics. The major difficulty comes from the large number of hair strands to be modeled, the fine geometries of individual hair, and the complex illumination interactions among the hair. Anew method, called the projection-based sweeping algorithm, is developed to render hair images using the cluster model. Compared with aprevious ray-tracing algorithm, the new algorithm offers several significant improvements: (1) it avoids the complex and expensive computation of the raycluster intersection; (2) it eliminates the time-consuming density-value evaluations at the sample points along aray during ray tracing; (3) it allows more flexible variation of the density map along the trajectory to enhance the natural appearance of hair; and (4) it allows easier coarse/fine control over the density map for multi-resolution rendering. As aresult, the new algorithm significantly improves rendering efficiency and hair image quality. Compared with the existing hardware-supported, texture-mapping projection methods, the new method allows more sophisticated shading models, accurate shadow calculation, and uniform handling of arbitrary slice orientations with accurate opacity accumulation.