Synthesis, manipulation, and rendering of volume-sampled geometric models
Sidney Wen-Hsin Wang · 1996
Rooted in the extensive research and development effort for visualization and analysis of sampled 3D volumetric datasets, volume graphics is beginning to attract traditional surface-based applications that deal with the modeling, manipulation, and rendering of synthetic scenes represented by geometric models. Some examples include terrain modeling for flight simulation, CAD modeling for shape design, and rendering of hypertextures, fur, and gaseous phenomena. Unlike the conventional surface graphics approach of representing the scene as a set of geometric primitives kept in a display list, volume graphics uses 3D volume buffers (3D rasters) as the medium for the representation and manipulation of a 3D scene. The scene is discretized in the preprocessing stage, and the resulting 3D discrete form is used as a database for manipulation and rendering purposes. Volume graphics offers several advantages due to the decoupling, uniformity, and atomicity features of the volumetric representation. It has the potential to revolutionize the field of computer graphics by offering an alternative to the existing surface graphics approach. This work details the advantages of volume graphics and contrasts each of the synthesis, manipulation, and rendering phases with conventional surface graphics. The synthesis phase includes proper voxelization techniques that incorporate signal processing theory to prevent object space aliasing, which is prominent in binary voxelization. A method for constructing a multiresolution hierarchy of volume rasters is developed. The multiresolution hierarchy is utilized during the rendering phase for speeding up the rendering process and for approximating penumbra shadow effects. The manipulation phase includes the creation of complex and realistic looking volumetric models through CSG modeling, where boolean operations are performed on a set of primitive objects; through volume sweeping, where the desired model is formed by sweeping a volumetric object along a trajectory in space; through volume morphing, where an intermediate volumetric model is interpolated from two 3D models; or through volume sculpting, where free-formed volumetric objects are created using the metaphor of sculpting a solid material. The rendering phase includes an extension of the volumetric ray tracing algorithm that capitalizes on the advantages of our voxelized models. By storing antialiasing information as part of the 3D volumes, alias-free 2D images can be efficiently generated without the need for image space supersampling. For speeding up the rendering process and reducing perspective depth aliasing, a hierarchical rendering approach is employed. In addition, the effect of penumbra shadows and fuzzy reflections are easily approximated using the multiresolution hierarchy in volume graphics.