One of the major problems in image processing is information reduction. The research reported in this thesis examines the semantics of cortext-independent edges in a real-world environment. Topics discussed include a classification scheme for edges, edge semantics, and a method for minimizing noise effects in edge computations.