This paper introduces a new representation system for three-dimensional shapes that is more succinct and natural than the commonly used generalized cylinders approach. Basic shapes are selected from a family of objects called superquadrics. This family contains all the standard modeling primitives, such as cylinders, spheres, and cubes, as well as the rounded shapes that are intermediate between these standard shapes. These basic shapes are thought of as “lumps of clay” that are subsequently deformed by, e.g., bending or tapering and combined by Boolean operations. Finally, 3D texture may be added to the surface of the objects using a fractal model. The motivation behind the use of superquadric modeling primitives, deformations, and fractal textures is that they seem to correspond well to human notions about shape and “part” structure.
An equation is derived that permits the identification of all important parameters of any family member from observing the tilt at a small number of points. This provides for an overconstrained representation, and additional points can be verified based on computed parameters. Common assumptions like smoothness, which seem hard to verify, are not made by the author. A solid modeling system called SuperSketch is described. It permits users to create fairly complicated objects much faster than with the usual CAD-type modeling systems.
The author’s paper is well written and combines references from an impressive number of scientific areas; thus, I only hesitantly note the following things that could be improved. Although there is a section on CAD modeling, there are no detailed references on that subject. It is not made clear how the parsing of a complicated object into “good parts” is intended to be done. Finally, it would be easier for the reader if examples for parameter instantiations for some members of the superquadrics family were given.