Abstract
We have previously shown that humans can discriminate changes in illumination and that sensitivity to such changes depends on the chromatic direction of the change and on the chromaticity of the surfaces in the scene. Here we explore the degree to which illumination discrimination taps a surface-independent representation. Specifically, we ask whether illumination discrimination remains possible when, unlike in our previous studies, the surface reflectances in the scene are shuffled as the illumination varies. Stimuli were rendered using RenderToolbox3 and viewed stereoscopically. On each trial subjects (N=10) viewed a scene illuminated by a target light followed by two test scenes — one illuminated identically to the target, another illuminated by a comparison light — and judged which matched the target. Across trials, the comparison light varied in chromaticity ("bluer", "yellower", "greener or "redder") relative to the target light (1-50 CIELUV ΔE units). The scenes were tiled with square surfaces, which were assigned one of 14 preselected surface reflectances; within a scene the area assigned to each reflectance was roughly equal. In one condition the surface reflectance assignment was fixed across scenes; in the other, the assignment varied randomly across the test and target scenes, while keeping the average scene reflectance roughly constant. With fixed surfaces, illumination discrimination was good and subject performance was similar to that measured in our previous studies. When the surfaces were shuffled, performance was significantly worse: for eight subjects, thresholds were systematically higher (by 14 ΔE on average); for the remaining two subjects, thresholds exceeded our largest step (50 ΔE) for at least one illuminant direction. Our results show that illumination discrimination is possible in the absence of fixed surface reflectance layout. Further research will attempt to determine whether subjects make the judgments based on a percept of scene illumination or whether they use task-specific strategies.
Meeting abstract presented at VSS 2016