Are edges sufficient for object recognition?

Abstract

The author s argu e that the concep t o f edges as used i n current research o n object recognition obscures the significant difficulties involved in interpreting stimulus information . Edges have sometimes been operationalized as line drawings, which can be an invalid and misleading practice. A new method for evaluating the utility of edge information, operationalized as the outputs of a local, signal-based edge extractor, is introduced. With 1-s exposures, the accuracy of identifying objects in the edge images was found to be less than half that wit h color photographs. Therefore, edge s are far from bein g sufficient fo r object recognition. Alternative approaches to the problem of interpreting stimulus informatio n are discussed. As with any complex process, initial research on object recognition ha s bee n conducte d withi n framework s that simplify som e aspect s o f th e process . Simplificatio n is important and necessary during initial stages of research. However, simplified frameworks must be elaborated upon or replaced on the road to a fuller understanding. In this article, we argue that the current concept of edge-based representations should be replaced because it is a simplification that obscures significant problems involved in the interpretation of stimulus information during object recognition. We will use Biederman's (1987) recognition-by-components (RBC ) model t o introduce ou r arguments because edge-based representations are central to it and because it has bee n a n importan t mode l o f objec t recognitio n i n psychology. However , simila r assumption s ar e mad e in