Abstract
In simple disk-annulus lightness displays, assimilation can influence the disk lightness over one range of annulus luminances, while contrast influences it over another range of annulus luminances (Rudd, 2010). To account for this pattern, I proposed a computational lightness model involving three neural processing stages: i) encoding of local oriented edge contrast; ii) contrast gain control between neurons responding to nearby edges; and iii) spatial integration of oriented contrast to compute lightness (Rudd, 2014, 2016). Previous experiments demonstrated that a luminance edge can be voluntarily included in this computation—or not—depending on the observer's interpretation of the edge as a reflectance edge or an illumination edge (Rudd, 2010). Here I present evidence for a second, distinct, type of top-down influence that controls the size of the spatial window over which spatial context influences the disk lightness. To explain quantitative lightness matching data with my model, the first type of top-down influence (edge classification) must occur early (i.e. pre-contrast gain control), while the second type (window size) must occur late (i.e. post-contrast gain control).
Meeting abstract presented at the 2016 OSA Fall Vision Meeting