ABSTRACT
AFTER HAVING WORKED THROUGH the important prerequisites, we are now ready to start with a first GPU-based implementation. We will first identify the individual components that a volume renderer is built upon. Afterwards, we examine different implementations and analyze their strengths and weaknesses both in terms of image quality and rendering performance.
We assume that the scalar volume is sampled on an equidistant rectangular 3D grid. This is the most common discrete representation of volume data in practice. All the techniques described in this chapter are objectorder approaches. The graphics hardware is used in the way it was designed for in the first place. We decompose the object into geometric primitives and then render these primitives using the GPU. Image-order techniques and the GPU-based ray-casting approach will be discussed in Chapter 7.
