The techniques presented in this work allow to visualize volumetric data at highly interactive frame rates even on standard desktop hardware. The key to high rendering performance is based on the observation, that in many applications of volume visualization the rendering of a small percentage of the volume's data is sufficient to depict the desired structures. The approach presented in this work -scanning for voxels which contribute to the visualization considering the chosen visualization parameters, and the extraction of relevant voxels into a secondary data structure - allows efficient skipping of non-contributing parts of the volume during rendering at zero cost. Depending on the chosen rendering methods and visualization scenario, this allows to get rid of up to 99% of the volume data without sacrificing spatial accuracy of the visualized objects. Although it is possible to perform high-quality, cell-based rendering using a similar data structure containing volume cells, the data structure is primarily designed to perform really fast rendering of single voxels, for example using a fast shear/warp-based projection. Clearly, the approach is a trade off between fast rendering and high quality of the images, as are also other interactive volume visualization techniques.
The explicit association of extracted voxels to segmented objects within the volume allows an extremely efficient handling of rendering and visualization parameters on a per-object basis. Without impact on rendering performance, color and opacity transfer functions, shading modes, and even compositing methods can be changed individually for each object. The flexibility of the approach in terms of the manipulation of visualization parameters, makes it well suited for data exploration and analysis.
The compact representation of the volume by the extracted voxels is well-suited for effective compression. It can be utilized to provide volume visualization capabilities over slow networks, like the Internet.
The combination of flexibility in terms of visualization parameters and high interactivity even on low-end hardware, makes the presented techniques unique.