Abstract
Defocus blur is a useful cue in many biological tasks. However, the extent to which the variability in natural images affects the discriminability of defocus blur is unknown. Here, we measured discrimination of retinal image blur created by the eye's optics, for a large collection of well-focused natural image patches. We constructed a psychophysical apparatus capable of presenting stimuli at three distances simultaneously along the line of sight. Half-silvered mirrors combined light from three monitors positioned at variable distances from the subject. The stimuli were rendered sharply on each monitor so that defocus blur was created by the optics of the subject's eye, as in natural viewing. Observers viewed stimuli through a 4 mm artificial pupil. Accommodation was paralyzed. On each trial, two different natural image patches were foveally presented in a 2IFC paradigm. The observer identified the interval with the sharper image. Throughout the experiment, each trial contained a different pair of image patches. Luminance was fixed across all trials (18 cd/m2). Contrast was matched within a trial, but varied between trials. Defocus blur discrimination thresholds were measured for three subjects across multiple standard defocus levels between 0.0 and 1.125D. For each standard, 6 comparison levels were presented, above and below the standard. The discrimination threshold was 0.18D on average across conditions. Thresholds did not vary significantly between subjects. Thresholds were constant below 1.0D, but rose for higher defocus levels. These findings suggest that natural image variability is a significant determinant of performance limits with natural stimuli under natural conditions.