Averaging illumination colors of multi-illumination ensembles

Average extraction from multiple similar items in ensemble perception has been demonstrated in a variety of visual features. This study investigated whether the visual system can extract precisely the average illumination color from multi-illumination ensembles. In a two-interval forced choice (2IFC) task of the constant stimulus method, observers were asked to judge whether the color of four identical test illuminations (homogeneous condition) or the average color of four distinct test illuminations (heterogeneous condition) was yellower than the reference illumination D67. Four types of scenes were used: Munsell surfaces only, Munsell surfaces with dark and light gray intervals, respectively, and an image of color patches on a white background. The mean values of the cumulative normal distribution functions fitted to the data under the heterogeneous condition were close to those under the homogeneous condition, indicating that the observer has the ability to precisely extract the average illumination color from a set of illuminations. This suggests that rapid and efficient averaging can also be performed in the high-level illumination estimation task.