Abstract
Veridical lightness requires veridical classification of reflectance and illuminance edges. When edges intersect, the luminance values in the four quadrants produce two basic patterns: ratio-invariance and difference-invariance. Difference-invariance occurs when two illuminance edges intersect. Ratio-invariance occurs when one reflectance edge and one illuminance edge intersect (but it does not reveal which edge is which). In Experiment 1, observers looking through a beam splitter saw two identical bright squares: a square veiling luminance and a square hole in a transparent filter. Forced choice judgments were random when the background was homogeneous. A square wave background produced X-junctions that were ratio-invariant at the square hole but difference-invariant at the square veil. Forced choice judgments were correct 100% of the time. This shows that X-junction information is effective. In Experiment 2, a large square piece of black paper was suspended in midair (in a dimly-lit room) and illuminated by light from a slide projector (Gelb effect), causing it to look like white paper. Ratio-preserving X-junctions were created by overlapping two regions on it: a white paper disk and a rectangular shadow (produced by a small opaque rectangle mounted on the glass projector slide). This display was viewed by a group of 15 subjects, who matched the square background to a Munsell 8.7. Of these, 14 reported the rectangle as black paper (Munsell 2.4) and the disk as a spotlight. A separate group of 15 subjects who viewed the same display surrounded by a thick white border saw the square background as black (Munsell 2.5). Twelve reported the black rectangle as a shadow and 11 reported the disk as white paper (Munsell 9.2). Overall, 25 subjects out of 30 reported an intersection of two kinds of edges, but the lightness of the large background determined which was which.
Meeting abstract presented at VSS 2013