One characteristic feature of RTVR is its way of handling and rendering of volume data, which is highly optimized to provide interactive feedback during the manipulation of viewing, rendering, and data mapping parameters. A high rendering performance is achieved by efficiently excluding non-relevant parts of the data from the rendering process. In common applications, like the visualization of medical data from CT or MR scanners, usually only a small portion of the data actually belongs to the object of interest. Furthermore, meaningful settings for rendering parameters may render the inner parts of objects opaque - a fact exploited by early ray termination techniques and also by the preprocessing included in RTVR.
RTVR interprets the input volume as being composed of objects, like bones, vessels, and other tissue making up, for example, a data set of a human hand (figure 6.1). The necessary segmentation information is either obtained together with the volume data itself from an external data source, or is interactively computed using simple threshold-based segmentation. For rendering, data mapping and rendering parameters can be individually assigned object by object. Besides the usual manipulation of object properties like opacity and color transfer function, also the shading model which is used for rendering can be defined individually for each object. This allows, for example, to combine objects which are rendered by using a standard shading model like Phong shading with objects that are rendered by the use of non-photorealistic shading [10,11]. In addition to the shading model, the way in which voxels are blended to the image plane can be defined in an object-aware way. Most volume rendering packages only allow to render a whole data set using either the usual opacity-blended compositing [26], or surface rendering [29,44], or maximum intensity projection (MIP) [42]. In RTVR, compositing modes can be selected on a per-object basis and combined with another inter-object compositing mode (two-level volume rendering [22]). This allows to choose the most appropriate rendering and compositing parameters for each object, depending on the structure of the data and the goal of the visualization.
![\includegraphics[width=.7\linewidth]{Figures/anmacher.ps}](img264.png)
|
Another feature of RTVR is to support the visualization of time series of volumetric data and of multi-dimensional parameter-series of volumes from simulation. The large memory demands of such data are compensated by the fact, that data extracted from a volume and used by RTVR for rendering is usually much smaller than the original volume. Only extracted data of the current volume has to be kept in memory for rendering, remaining parts of the volume and data which belong to other time (parameter) steps are placed into a combined memory/disk cache.
Among other ``standard'' features of volume visualization systems, RTVR provides the ability to display data on planar sections through the volume, enables object and position picking by clicking into the rendered image, and supports the clipping of volumes or sets of individual objects at planes and more complex structures. Data which has been ``clipped away'' can be omitted from rendering - which is the most common approach - or rendered using different rendering parameters, for example, more transparent than the non-clipped part of the object, or using a different shading model, which for example, displays just contours.