Abstract
Lightness matches made to papers in Staircase Gelb and modified Staircase Gelb displays were modeled with a neural model of lightness computation based on the principle of edge integration. To compute lightness, the model sums logarithmic steps in luminance at the edges of all papers in the display along paths directed towards the target paper, with edge weights that depend on two independent factors. Factor 1 is the distance of an edge from the target. Factor 2 depends on whether the luminance step at the edge increments or decrements in the target direction. Factor 2 was estimated from quantitative fits to ON- and OFF-cell responses in macaque LGN (De Valois, Abramov, & Jacobs, 1966; Billock, 2018). In the perceptual experiments, grayscale papers were arranged either: 1) from lowest to highest reflectance in a spotlight (Zavagno, Annan, & Caputo, 2004, Series A); 2) with the papers spatially reordered such that the highest reflectance paper neighbored the lowest reflectance paper (Zavagno et al., Series B & C); or 3) as in (1), but with an white border surrounding the display (Gilchrist & Cataliotti, 1994). The neural model reproduced to within 1.6% error the average lightness matches made in these experiments. It accounted for both the overall magnitude of dynamic range compression observed when the papers were spatially well-ordered, and various releases from compression that were observed when the papers were spatially reordered. It also reproduced the observation of Gilchrist and Cataliotti that surrounding the display with a white border brings the perceived reflectance scale more in line with the true ratio scale of the physical paper reflectances (ground truth). These results demonstrate that a plausible neural model of lightness, having parameters derived from physiology, can explain the quantitative lightness scaling of real material surfaces viewed under spotlight illumination in these experimental conditions.