Volume visualization has been a fast growing area of research for over two decades. Its objective is to enable viewers to understand three dimensional data sets. Volume visualization has been driven by the advances of medical image acquisition, that was induced with the appearance of computed tomography (CT), magnetic resonance (MR), positron emission tomography (PET) scanners and other imaging devices. These machines produce three dimensional arrays of digital sampled data by imaging a series of cross section that represent spatial volumes of a specific substance property. An MRI scan for example illustrates well the location of soft tissue, especially tumorous ones.
Usually, projection needs to be applied on the numerical description of a volume, to extract easy to understand information. The three dimensional data set is projected onto a two dimensional image plane in order to see the underlying structure that is contained in the volumetric data. These volumetric data are hard to effectively visualize since three dimensional structures of the interior of a volume are difficult to derive from viewing individual slices. A better three dimensional perception is given by integrating these slices together in a volume that can be seen from any viewpoint and containing lighting effects that reinforce surfaces, as well as small variations in density and opacity.
Volume rendering has been one of the main techniques to visualize the exterior surface and the inside of the volume data that these medical scanners provide. Marc Levoy invented the first volume rendering algorithm (Lev88). It is based on a simple model of light interaction with surface density and opacity within the visualized volume. The impact of light with optical density is calculated by following conventional models of light transport. His work has been subsequently extended, adding features and improving performance.
This project concentrates on the earlier literature of volume rendering, especially, Levoy's seminal contribution. Despite the simplicity of his underlying model, there are lots of complications that are often understated. It is hard to systematically obtain satisfactory results and sometimes impossible to achieve good visualization because of features like noise contained in a specific data set. This project goal was to create from first principles a set of tools that enable many different variations on Levoy's basic idea. This makes it possible to explore the basic factors that make one visualization successful and another unsuccessful.