Foundations of Vision

Abstract The ease with which we can perceive and recognize things from a brief glance belies the staggering complexity of human vision. Here, I will describe the computational challenges faced by the visual system and current understanding of its neurocomputational workings. Psychophysical investigations, informed by introspection, have provided critical windows into the nature of vision. Theories of information processing such as signal detection theory and Bayesian inference provide conceptual frameworks for understanding perception and its limits. The functional architecture of the visual system is reviewed to clarify how information is transformed across successive stages of the visual pathway. With this knowledge in hand, we will consider the neural mechanisms that underlie the perception of basic features and complex objects, and the role of attentional feedback in optimizing perceptual processing. This review highlights recent advances, including biologically inspired models of deep learning, that establish our growing understanding of the neural computations that mediate visual perception, attentional selection, recognition, and inference.