Explorative volume rendering using the Max-Tree datastructure

When working with volume data such as CT or MRI scans of the human body, it is often challenging to interpret the data. The volume is projected on a two-dimensional image, often resulting in overlapping elements in the image. Techniques have been developed to make it easier, like having an interactive view, looking at 2D slices through the volume, using various colors schemes and applying filters to the data. There are many methods to select components in the volume data and color the volume, but the size of volume datasets makes it hard to use them interactively. The Max-Tree datastmcture makes it possible to experiment with new techniques to explore volume data. It allows more speed operations dealing with connected components, making it possible to use them interactively. This thesis is about explorative volume rendering. The goal of my research was to develop a tool for explorative volume rendering based on the Max-Tree datastructure representation of data. In explorative visualization, the user plays an active role in the way data is represented. The user can change selections, parameters or viewpoints to get a better insight in complex datasets. With volume rendering there is the problem that a volume is rendered to a two dimensional image. Some details might be hard to spot without removing parts of the volume that obstruct the view. This research looks at ways to modify the volume dataset based on the Max-Tree data structure. This structure holds a decomposition of a volume dataset, and allows modifications without changing the structure of the objects inside. The structure allows selections of components based on shape or size, making it easy to remove or select parts with certain features of a volume. Using data structures for explorative visualization was shown useful before with contour trees [2], where a network of contours at different isovalues is built. The Max-Tree data structure can also be used to construct various transfer functions. These functions give colors or opacity values to the voxels in the volume, to create more clear rendering of a dataset. There are many attribute setting that can be used to determine parts that should belong together, and have their own color. Part of this research will be to find some useful ways to construct these transfer functions based on the structure and attributes of the dataset. This report starts with some background information about the subjects that were important for the research. What follows are the actual research subjects, about the rendering of the Max-Tree data structure and its interaction and about the construction of transfer functions. It will end with a discussion about the results.