This paper proposes a novel algorithm for surface-based hierarchical clustering simplification that aims to accelerate view-dependent point set rendering.The remarkable advantage of this algorithm is that it uses a normal cone semi-angle metric to trace surface curvature variation and provides the clustering simplification process with a reliable global error control

rather than commonly used spatial partition scheme.In off-line simplification phase

combined with various predefined clustering constraint conditions the algorithm constructs a continuous multi-resolution hierarchical representation for point set model.In real-time rendering phase

hierarchical visibility culling and optimized tree traversal efficiently improve system integral performance.Moreover

an additional silhouette enhancement mechanism is introduced to ensure a well rendering vision quality in spite of a larger screen projection error and higher model reduction rates.