We live in a complex visual environment and do not process all of the available visual information to the same degree. Our visual system has a finite processing capacity and needs to ensure that our behavior remains efficient with regard to what we are currently doing. Processes of visual attention select behaviorally relevant visual information for perception, action, or memory storage and suppress processing of irrelevant information.
Eye movements are closely connected to visual attention and constitute a key component of real-world vision. We constantly move our eyes whenever we search for particular objects, examine our environment, or have to decide whether a visual scene is familiar or not (e.g., when we explore a foreign city and have to decide where we are and where to go).
Fast abrupt eye movements (i.e., saccades) bring different parts of a scene (or an image) to the center of our view, where acuity is highest. During periods of stable gaze (i.e., fixations) our visual system gathers detailed visual information about the object, or location one is looking at. When single objects or people pass through a scene we often follow them with our eyes (i.e., we perform smooth pursuit eye movements).
Eye movements reflect which visual characteristics attract attention. Analyzing the direction of fixations, saccades and smooth pursuit eye movements and relating them to the properties of the visual stimuli opens a window to human attentional processes.
In the present research project we aim to develop and empirically test a sparse, realistic, and ecologically valid model of human visual attention and gaze behavior in dynamic visual environments. We test our model in controlled psychophysical experiments and evaluate its predictions in applied settings, particularly in the context of biomedical dynamic visual displays as well as media coding and transmission.
Our research efforts should not only be relevant to basic research in psychological vision science and computational vision but also raise implications for the development of the next generation of visual applications in communication, workplace, and entertainment technology.
Different to previous computational models of visual attention our approach flexibly adapts to individual preferences and past visual experience of human observers. Furthermore, our model also accounts for visual motion and optic flow as major features of human vision under real-world conditions (where we constantly shift our eyes, heads, and viewpoints).
The recording and analysis of eye movement behavior (using video-based eye trackers) is of central relevance to our research. We also test some assumptions of our model using functional magnetic resonance imaging (fMRI) and electrophysiology (EEG/ERPs).