Abstract
The amplitude spectrum of natural scenes falls inversely with spatial frequency (f) according to 1/fa, where a dictates the rate of descent in amplitude, ranging between 0.6 and 1.6 in natural environments. Human observers are sensitive to the value of a: discrimination thresholds peak when reference a ranges between 1.0 and 1.3. This peak in sensitivity is thought to be related to the prevalence of as values encountered in the natural world: most scenes have as of approximately 1.0 (Hansen & Essock, 2005, Vis. Cog.). Indeed, an adaptation study (Elliott et al., 2011, JoV) has shown that adapting to shallower as makes shallow a images appear steeper, while the opposite holds true when adapting to steeper as. These effects are indicative of a renormalization to some pre-adapted state of the human visual system. If this is true, then placing observers in a natural environment with a values that deviate from the 1.0 average should shift the peak in a sensitivity towards the new environment average. We investigate this using modified reality: we immersed 3 observers in an environment where the a of images in either steepened (a + 0.4) or shallowed (a - 0.4). Observers were placed in modified reality for a total of 60 minutes and a discrimination thresholds were measured for 5 reference a values (0.4, 0.7, 1.0, 1.3, 1.6) at 20 minute intervals. Steepening a resulted in flat tuning to a for the first 20 minutes of adaptation while shallowing a had little effect. In the steeper environment, thresholds for all reference a returned to baseline values, except for a reference a of 0.7 following adaptation (60 min). We explain these effects with a model of contrast adaptation that pools the responses of multiple, differently tuned spatial frequency channels, according to the 1/f property of natural scenes.
Acknowledgement: NSF CHS-1524888