Abstract
When identical achromatic test disks are placed on grayscale images of natural scenes, the disks placed in low luminance portions of the image appear brighter than the disks placed in higher luminance portions of the image. Shapiro and Lu (2011) demonstrated that the relative brightness can be accounted for by removing low spatial frequency content from the image. Here, we examine whether the visual system filters the amount of low spatial frequency content in terms of object units (i.e., relative to the size of the disks) or in terms of retinal units (i.e., set at a fixed visual angle). Observers ranked the perceived brightness of seven physically identical test disks placed on grayscale images of natural scenes. Five disk diameters were used (20, 40, 80, 120, or 160 pixels), and each observer viewed the images from four distances (60, 100, 200, 300 cm). A single parameter model controls the amount of low spatial frequency content removed from the image (cut-off frequency); we report the correlations (r) between observer rankings (perceptual values) and rankings of the disks’ pixel (i.e., physical) values as a function of the cut-off frequency. The correlation values typically peak with r > 0.85 when the cut-off frequency is at or greater than the pixel size of the disks; for each disk size, the peak correlation occurs at the same filter level regardless of viewing distance. Perceived brightness of objects embedded in natural scenes is therefore independent of viewing distance, based not on the absolute size of objects on the retina, but on their size within the image. Our results suggest that brightness perception involves scale invariant computations of object properties, from which the amount of low spatial frequency content filtered from the image is derived.
Meeting abstract presented at VSS 2012