Voxels versus Polygons: A Comparative Approach for Volume Graphics

The decision to use either the Direct Volume Rendering paradigm or the Indirect Volume Rendering paradigm to visualize a volume dataset is a topical question in Volume Graphics. Unfortunately, it seems that this question has not been sufficiently addressed so far and is not easy to answer. In this paper, we discuss some of the advantages and disadvantages of one candidate for each of these paradigms; Ray Casting for Direct Volume Rendering and Marching Cubes for Indirect Volume Rendering. The discussion is based on two cartesian grid scalar data fields, a CT scan of a lobster – immersed in resin, and a MRI scan of a human head. These datasets have interesting properties which show different features using different Volume Rendering techniques. Two measurements are considered for our discussion; visual quality and the consumption of resources, such as time and memory.