Abstract: With the increasingly sophisticated demands in practical applications, challenges emerge in expressive rendering respect to computing efficiency, data throughout and rendering quality. Aim at visualization applications for large-scale 3D data, this paper proposes an efficient framework of interactive expressive rendering. First, based on Map Reduce, we propose an adaptive method which parallelly rasterizes 3D models in high resolution. Then, a data storage schema named interactive pixel block（IPB） is designed. Based on the schema, we propose a high resolution expressive rendering solution suitable for online interaction, in order to achieve all kinds of rendering effects including non-photorealistic shading, silhouette rendering and transparent rendering. Finally, we realize a light-weight interactive visualization service in B/S architecture. For any 3D model, the method proposed in this paper only need to rasterize it once in order to provide the online service, which enables users to view all kinds of expressive rendering results and to interactively explore those large-scale 3D models. According to the results, this method can be used in many large-scale visualization applications such as large-scale CAD design, interactive demonstration in museums and scenic spot instruction.