Thanks to the assistance of efficient graphics hardware, the Projected Tetrahedra algorithm [86] was the fastest known unstructured-mesh rendering algorithm for over a decade. The view independent cell projection algorithm, developed by Weiler and colleagues [98,100] and reviewed in Section 1.1, improves on the Projected Tetrahedra algorithm by taking advantage of graphics hardware capabilities that were not available when Projected Tetrahedra was developed.
In Section 1.2, I proposed changes to the view independent cell projection algorithm that reduce the demands of the bandwidth between CPU and GPU. We shall see in Chapter 1 that these changes greatly improve the speed of the algorithm.
In Section 1.3, I added adaptive transfer function sampling to the algorithm. It divides the tetrahedra such that material properties vary linearly within each piece. This linear variance makes our integration of volume properties (discussed in the next chapter) far more accurate. This tetrahedron division ultimately slows down the algorithm. The changes send more data to the card and add tetrahedra to the rendering. However, we require better transfer function sampling such as this to achieve high quality renderings. We shall see the overall effect of adding adaptive transfer function sampling in Chapter 1.