Identifiable Images of Bystanders Extracted from Corneal Reflections (Extract)

Previous psychological research has established that humans can identify faces from extremely poor quality images, when they are familiar with the faces concerned[4]–[7]. For example, Yip & Sinha[7]found that viewers could identify blurred photographs of familiar faces with equivalent image resolutions as low as 7×10 pixels (see Figure 1). Here we exploit the robustness of familiar face recognition to mine high-resolution portrait photographs for latent information. Specifically, we show that the faces of hidden bystanders can be identified via reflections in the eyes of photographic subjects. Corneal analysis has previously been used to recover coarse aspects of the physical environmental, such as ambient lighting conditions[8][9]. The present findings demonstrate that corneal reflections can reveal surprisingly rich information about the social environment too.