Abstract
To maintain color constancy, our visual system must distinguish surface reflectance-based variations in wavelength and luminance from variations due to illumination. Edge integration theory proposes that this is accomplished by spatially integrating along an image path only those local oriented luminance and color contrasts that likely result from reflectance changes. The output of the edge integration process is a neural representation of relative reflectance across the visual scene. An anchoring rule — the largest reflectance in this representation appears white — maps the representation of relative reflectance onto an absolute lightness scale. Computations occur in this order: edge classification -> weight setting -> edge integration -> anchoring -> lightness. Lightness matching data from several studies using 2D displays are shown to be consistent with an edge integration model in which the visual system performs a weighted spatial summation of directed steps in luminance, where luminance is defined in log units. Three hypotheses are proposed regarding how weights are applied to steps. First, weights decline with distance from the surface whose lightness is being computed. Second, larger weights are given to steps whose dark sides point towards the surface. Third, edge integration is carried out along a path leading from a common background, or surround, to the surface. The last rule is needed to make the other two rules work in a self-consistent manner and implies an indispensable role for figural organization in any viable edge integration theory of lightness. The theory accounts for: simultaneous contrast; quantitative lightness judgments in disk-and-annulus, Gilchrist dome, and Gelb displays; and perceptual filling-in. I explain the motivation for the model and how it accounts both for 1/3 power-law lightness scaling of Gelb papers when the papers are surrounded by darkness and the veridical lightness scaling that occurs when the same papers are surrounded by a white frame (Cataliotti & Gilchrist, 1995).
Meeting abstract presented at VSS 2013