Abstract
Object recognition concerns itself with two questions: What is the form of object representation? and How do observers match object percepts to object representations? Many objects look similar and most contain no single feature that uniquely identifies them. Furthermore, objects are rarely seen under identical viewing conditions: Objects change their size, position, orientation, and relations between parts, viewers move about, and sources of illumination turn on and off or move. Successful object recognition requires generalizing across such changes. Two different approaches to these issues have been adopted. Viewpoint‐invariant theories assume that there are specific invariant cues to object identity that may be recovered under almost all viewing conditions. Viewpoint‐dependent theories suggest that no such general invariants exist and that object features are represented much as they appeared when originally viewed, thereby preserving shape information and surface appearance. Despite many differences, theories of object recognition include some common principles. These include the decomposition of an image into component features, the coding of the spatial relations between such features, multiple views to represent feature sets arising from different object viewpoints, generalization mechanisms to normalize over changes in viewing conditions, and the flexibility to support recognition tasks ranging from item‐specific individuation to basic‐level categorization.
