One of the main problems in volume rendering is context. It is hard enough to find a transfer function that shows the inteersting parts in the volume data, but finding one that shows the whole volume as context and the interesting in a way that they are not occluded by other parts of the volume is very hard. At IEEE Visualization 2004 a group from the Technical University of Vienna published a paper called Importance-Driven Volume Rendering that promises a solution to this problem by incorporating importance into the process.
This importance is used to do a segmentation on the dataset that builds a geometric representation of the important parts of the data. This geometric representation is then used to cut out parts of the volume that could occlude the important ones. The problem with approach is the need for segmentation, which is a time-consuming process and therefore precludes interactive exploration of the volume, which is important for volumes where the contents and the goals are not initially known. The goal of this project is to build a system that uses importance at the same level as the general lookup table, allowing much more interactive changes and explorative use of the volume data.