Dual mechanisms of curvature detection

Abstract

Purpose: The detection of object curvature is an essential feature of human spatial vision. We investigate the combined effects of contrast and blur on thresholds for curvature detection.Methods: Horizontally‐oriented line stimuli were displayed on a Macintosh monitor. The lines were windowed in both x‐ and y‐dimensions with a Gaussian profile. We examined a range of x‐ and y‐blur levels by varying the space constant of the relevant Gaussian window, such that stimuli varied in appearance from long, thin lines, to short fat ones. The maximum contrast of the lines varied between 12.5–100%. Subjects were required to indicate whether the line (presented for a duration of 250 msec) was curved upwards or downwards. A method of constant stimuli was used to determine curvature thresholds.Results: At each level of blur (y‐space constant), threshold remained constant up to a certain line length (x‐space constant) after which they increased. Thresholds also increased with increasing levels of blur. However, performance in all conditions became identical once thresholds were expressed in Weber terms (as a multiple of line length) and stimulus size was specified using a scale‐invariant metric (ratio of line length to blur). When line length is small relative to blur (short, fat lines), thresholds are proportional to blur and are contrast dependent. When line length is large relative to blur (long, thin lines), thresholds become proportional to line length and demonstrate reduced contrast‐dependence.Conclusion: Combined manipulation of blur and contrast reveals two distinct processes involved in curvature in human vision.Acknowledgements: Supported by the Visual Research Trust and a Wellcome Trust Vacation Scholarship to JAD.