Visual object representation: An introduction

What are the computational, behavioral, and neural mechanisms that give rise to object perception? In this review, I present a cognitive neuroscience overview of the literature on object representation. Marr's (1982) framework for studying complex tasks is used as a guide for the review. This framework involves analyzing a problem on three levels: (1) the computational theory, which asks what is computed and how; (2) the representation and algorithm, which focus on the representations and processes that underlie a computation; and (3) the hardware implementation, which deals with the implementation of the representations and processes. Computational considerations of object recognition raise the importance of the object invariances, which allow viewers to perceive an object as remaining stable despite changes in the retinal image. I then use the invariances to guide my review of the representations and processes involved in human object recognition, Marr's second level, and of the hardware implementation, Marr's third level. Throughout the review, my focus is on integrating across disciplines and across the levels of Marr's framework.