Abstract
The scientific simulation and three-dimensional imaging systems in use today are producing large quantities of data that range from gigabytes to petabytes in size. Direct volume rendering, using hardware-based three-dimensional textures, is a common technique for interactively exploring these data sets. The most serious drawback of this approach is the .nite amount of available texture memory. In this paper we introduce a hybrid volume rendering technique based on the use of hardware texture mapping and point-based rendering. This approach allows us to leverage the performance of hardware-based volume rendering and the flexibility of a point-based rendering to generate a more efficient representation that makes possible interactive exploration of large-scale data using a single PC.