Title: Visual Identification and Search in Natural Scenes
This talk will describe a new principled theory of visual search developed within the framework of natural scene statistics and Bayesian statistical decision theory. The theory is unique in several ways: (1) it directly takes into account the statistical properties of natural background images, (2) it takes into account the variation in neural processing with retinal location (i.e., the limitations of peripheral vision), (3) it takes specific images as input and produces perceptual decisions and fixation locations as output, (4) it is strongly constrained by measurable properties of the stimuli and known properties of the visual system, and hence contains almost no free parameters, and (5) covert (single fixation) and overt (multiple fixation) search are predicted within the same coherent framework. The theory is developed for search in natural scenes, but makes detailed quantitative and qualitative predictions for arbitrary specific search stimuli. The presentation will illustrate for students a number of basic computational principles of human vision.