Abstract
Observers exhibit lightness constancy both when the illumination level on a scene is increased and when a homogeneous layer of light (a veiling luminance) is added to the scene, even though these transformations of the retinal image are very different and no lightness theory can explain both. When a Mondrian pattern is seen through a veiling luminance without visible borders, patches on the Mondrian appear much lighter gray and the veil is not perceived; zero constancy. But when the same veil covers a 3D scene of equivalent luminance range, lightness constancy is almost 100% and the veil is perceived. What information is used to reveal the veil and estimate its strength? We report nine experiments using a new apparatus that allows a light source to be placed either at the virtual location of the eye (thus eliminating both cast and attached shadows) or 25 degrees away from the eye. We found: (1) Lightness constancy is significantly better (blacks look blacker; veil perceived) when shadows are present. Adding a veil creates a positive correlation between the contrast at a shadow boundary and the luminance of the surface it falls on. Without the veil there is no such correlation; all shadow boundaries have equal contrast. (2) Adding colored patches to the Mondrian had no effect while adding colored objects to the 3D scene produced better constancy. In this case, adding a veil creates a positive correlation between the saturation gradient across a curved colored object and the luminance gradient across it. (3) A weakly colored veil (24% purity) produced better constancy than a neutral veil. In this case the veil creates a negative correlation between the saturation gradient across a curved achromatic object and the luminance gradient across it. Constancy was significantly lower when the curved objects were replaced by rectangular objects.
Meeting abstract presented at VSS 2016