Identifying perceptually significant features for recognizing faces

Abstract

The human visual system possesses a remarkable ability to detect and identify faces even under degraded viewing conditions. The fundamental challenge in understanding this ability lies in determining which facial attributes the visual system uses for these tasks. Here we describe experiments designed to probe the limits of these abilities and determine the relative contributions of internal versus external facial features for the detection and identification tasks. The results provide strong constraints and guidelines for computational models of face perception.