Semi-automatic transfer function generation for non-domain specific direct volume rendering
The field of volume rendering is focused on the visualization of three-dimensional data sets. Although it is predominantly used in biomedical applications, volume rendering has proven useful in fields such as meteorology, physics, and fluid dynamics as a means of analyzing features of interest in three-dimensional scalar fields. The features visualized by volume rendering differ by application, though most applications focus on providing the user with a model for understanding the physical structure represented in the data such as materials or the boundaries between materials. One form of volume rendering, direct volume rendering (DVR), has proven to be a particularly powerful tool for visualizing material and boundary structures represented in volume data through the use of transfer functions which map each unit of the data to optical properties such as color and opacity. Specifying these transfer functions in a manner that yields an informative rendering is often done manually by trial and error and has become the topic of much research. While automated techniques for transfer function creation do exist, many rely on domain-specific knowledge and produce less informative renderings than those generated by manually constructed transfer functions. This thesis presents a novel extension to a successful semi-automated transfer function technique in an effort to minimize the time and effort required in creation of informative transfer functions. In particular, the method proposed provides a means for the semi-automatic generation of transfer functions which highlight and classify material boundaries in a non-domain specific manner.