This project explores the nature of visual scene representations through behavioral and computational work. Though perception of scenes and visual cognition have been studied in isolation, no computational architecture has been proposed that might integrate perceptual phenomena with visual reasoning and inference. Here, it is suggested that a representational scheme based on functional relations would provide an elegant link between scene recognition and the processes supporting inference about actions and goals appropriate to that scene. Initially, it is hypothesized that a functional representation of visual scenes will have observable effects on attentional mechanisms of vision, in that functional groupings within a visual scene may be selected by visual attention. Additionally, it is hypothesized that the development of functional representations will influence a participant?s ability to detect important changes in a visual scene. In addition to two experimental approaches, a computational exploration (in the form of an explicit implementation of this theory) is proposed. In total, this research has potential to expand our understanding of the way people reason about their visual environment, and to help to refine training and educational techniques.
Green, Collin; Hummel, John E (2006) Familiar interacting object pairs are perceptually grouped. J Exp Psychol Hum Percept Perform 32:1107-19 |