In computer graphics, two types of geometric representations are widely used: surfaces and volumes. Surfaces are infinitely thin skins defining the boundary of an object, and have been successful in modeling realistic appearances. However, their lack of information about the solid nature of objects make them inadequate for many applications. Volumes represent the entire solid geometry of an object in a three-dimensional grid. Volumes, however, are inefficient and cannot achieve high resolutions since they must maintain information about internal geometry that is not pertinent to rendering realistic appearances.
We have created a new, hybrid data structure, the volumetric surface, which combines the benefits of surfaces and volumes. A volumetric surface represents solid materials in a thin region near the surface of an object in an efficient, multi-resolution data structure. Volumetric surfaces are thus designed to exploit the coherency of roughly surface aligned-layers of materials which are ubiquitous in real-world materials. Examples of such materials include wood, stone, and building facades.
To demonstrate the utility of volumetric surfaces, we present a system for the interactive sculpting of weathering and erosion effects on building facades. High-resolution sculpting of layers of solid brick, plaster, and mortar are not possible with surfaces or volumes alone. Volumetric surfaces are well suited to the representation and intuitive editing of such complex appearances. This thesis presents detailed algorithms for the sculpting system with a focus on efficient rendering and manipulation of volumetric surfaces.
The above images were interactively sculpted on simple brick wall in real-time. For a much more detailed discussion of this project, please download my Masters thesis. You can also email me for much higher resolution images, or a Postscript version. This work was done in the MIT Computer Graphics Lab under the supervision of Professors Julie Dorsey and Leonard McMillian.