Abstract
Because of factors such as the eye's limited depth of focus and accommodative lags, blur in the retinal image varies continuously over time and space. We examined how these variations are integrated in visual adaptation to image blur. Grayscale images were composed of a dense and random collage of rectangles and were filtered by varying the slope of the log amplitude spectrum over a range from −1 (strongly blurred) to +1 (strongly sharpened) relative to a focused slope of −1, with rms contrast held constant. The images subtended 4 deg and were varied in time to avoid local light adaptation. Subjects adapted for 60 sec to image sets drawn from a single slope or to a set of hybrid stimuli formed by combining pairs of images with different slopes. Test images were then interleaved with adaptation top-ups while the blur level was adjusted with a 2AFC staircase to determine the level of subjective best focus. Adaptation to blurred images caused physically focused images to appear too sharp and vice versa. These shifts were weaker in the hybrid images, suggesting that the adaptation is determined by the average image blur and not by the sharpest feature; but were biased toward sharper slopes, suggesting that this average is not linear with slope. In a second set of measurements, hybrid adapting sets were created by instead alternating in time between blurred and sharpened slopes. These again produced intermediate aftereffects relative to a single adapting slope. Measurements of the alternation rates at which the aftereffects become phase-dependent (ie dependent on the last presented slope) provide a measure of the integration time controlling blur adaptation, while combinations of different slopes within this time provide an estimate of how different blur levels are weighted to set the observer's state of adaptation to blur.