Visual adaptation to natural scene statistics and visual preference

The amplitude of Fourier spectra for natural scenes falls with spatial frequency (f) and is described by the equation, 1/f^α, where exponent α corresponds to the slope of the spectral drop-off. For natural scenes α takes on intermediate values ~1.25, reflecting their scale invariance. It is also well-established that, on average, images with natural scene statistics are preferred to those that deviate from these properties. Although this average pattern of preference for images with the intermediate values of α is robust, there are also marked individual differences in preference for different levels of α. This study investigated the effects of adaptation on average and individual visual preferences for synthetic filtered noise images varying in α. Participant preferences (N = 58) were measured via a 2AFC task prior to adaptation (baseline) and post-adaptation There were 3 adaptation conditions (α = 0.25, 1.25, 2.25) and 5 test levels of α (0.25, 0.75, 1.25, 1.75, 2.25). On average, the adaptation elevated preferences for test images with α matching the adaptor conditions, especially in adaptor conditions, α = 0.25 and 2.25. We also observed marked individual differences in preference for different levels of α. These different preference profiles remained stable throughout the experiment and affected the levels of adaptation observed in different adaptation conditions.