Geons

Geons are the basic volumetric components proposed in the Recognition-by-Components (RBC) theory of object recognition, introduced by Irving Biederman in 1987. In RBC, complex shapes are understood as assemblies of a finite set of geons, which are assumed to be invariant to viewpoint and lighting. Geons are simple 3D geometric shapes—such as cylinders, cones, spheres, wedges, and prisms—or simple combinations of these shapes. The theory posits a finite set of geons that can be combined and arranged to form the variety of everyday objects.

Recognition proceeds by extracting the edges and other cues from an image, identifying the geon constituents, and interpreting their spatial relations. Because geons are defined by non-accidental properties—features that tend to persist across viewpoints, such as curvature, parallel edges, cotermination of edges, and symmetry—the resulting object description is relatively robust to changes in viewpoint, occlusion, and lighting.

Strengths and criticisms: RBC accounts for rapid recognition of many man-made objects from partial or degraded information and has spawned extensive research on structural descriptions of vision. Critics note that many natural objects rely on texture, shading, color, and detailed surface properties not captured by a limited geon repertoire, and that real-world recognition may require more than a fixed set of primitives.

Impact: Geons influenced cognitive psychology, computer vision, and neuroscience by proposing a tangible mechanism for viewpoint-invariant object representation and by guiding experiments on structural descriptions of scenes and objects.

---