Phase-specific interactions in the perceived blur of edges

Abstract

The perceived blur of an image can be strongly affected by prior adaptation to a blurred or sharpened image or by induction from a blurred or sharpened surround (Webster et al., Nature Neuroscience 2002). We examined some of the stimulus properties controlling these interactions by comparing the adaptation and induction effects in simple edges with different luminance profiles but equivalent amplitude spectra. Stimuli were vertical edges, with amplitude spectra filtered over a range of slopes relative to 1/f. Different edge profiles were formed by shifting the phase of the component frequencies by fixed increments of pi/4. A staircase procedure was used to adjust the spectral slope until the edge appeared properly focused. For phases other than 0 and pi, the 1/f edges appeared blurred and therefore had to be physically sharpened. These perceptually blurry 1/f edges might therefore be expected to induce sharpness in a squarewave test, yet adaptation to them instead caused the test to appear blurrier, suggesting a dissociation between perceived focus and the effective focus for the adaptation. The magnitude of the adaptation and induction was also selective for the relative phase relations of the components and strongly selective for the contrast polarity of the edges. These results suggest that the perceived blur and blur interactions in local edges cannot be accounted for by the relative amplitude of the different frequency components and may instead depend on (potentially different) properties of the spatial luminance gradients of the edges.