Visual search has typically been studied using button-press dependent measures and fairly simple stimuli, methodological choices that have rendered current theories of search poorly equipped to predict the locations and durations of individual search movements to featurally complex real-world objects. In response to these limitations, the goals of the proposed project are to: (1) describe real-world visual search in terms of directly observable and spatio-temporally exact eye movement behavior, and (2) introduce a computational model capable of accommodating real-world oculomotor search. Work on this interdisciplinary project will be accomplished in three phases. Phase I will implement the computational model. Filter-based image processing techniques will be used to represent the real-world search stimuli, and visual routines acting on these representations will endow the model with simulated oculomotor behavior. The visual information available to each of these eye movements will be constrained by a simulated fovea that moves over the scene as the model's """"""""eye"""""""" gradually converges on the search target. Phase 2 will apply this behavioral and computational approach to address basic questions regarding real-world oculomotor search (set size effects, target presence/absence, etc). Behavioral studies will determine how people direct their gaze as they search for simple and real-world targets. Computational studies will then input to the model the same search scenes viewed by the human observers and compare the simulated eye movement behavior to the sequence of saccades and fixations obtained from the behavioral studies. Phase 3 will build on the results of Phase 2 by using the gaze patterns predicted by the model to test several novel questions regarding search processes and representations (e.g., What is the role of complex backgrounds in a search task? Should search items be treated as objects or spatially extensive image patches?, What are the relationships between visual search and memory?). Finding spatio-temporal agreement between human and simulated gaze patterns in these studies will not only provide the literature with a validated computational model of oculomotor search, but also open to researchers a (real) world of stimuli to challenge our understanding of visual search behavior.
Showing the most recent 10 out of 31 publications
