Pamela Reinagel and Anthony M. Zador
Sloan Center for Theoretical Neuroscience
Caltech and Salk Institute
Early stages of visual processing may exploit the characteristic low-level statistics of natural image ensembles. Animals actively select visual stimuli by orienting their eyes. In humans, eye positions determine which portions of a visual scene will fall on the fovea and thus be sampled at high spatial resolution. We recorded the eye positions of human subjects while they viewed natural scenes. We report that active selection changes the low-level statistics of the ensemble the visual system encounters. We express this difference in terms of a conventional spatial-frequency analysis, and also in terms of a wavelet-based analysis which indicates that subjects look at informative image regions.