A cue-free method to probe human shape from shading

People readily perceive patterns of shading as 3-D shapes. Owing to the generalised bas-relief ambiguity when extracting shape from shading, people must simultaneously estimate the shape of the surface and the nature of the light source. In many cases cues in the image will be insufficient to resolve all of the ambiguities present, and in such cases the human visual system may employ one of a number of prior assumptions based on ecology and experience. One such assumption is the lighting-from-above prior. Here, in the absence of extrinsic cues to lighting direction, ambiguous shading patterns are interpreted as if lit by a light source that is above the observer’s head. Studies of this prior typically use ambiguous stimuli and observe perceptual biases. A degree of cueing is inherent to such methods. Participants see the shaded stimuli repeatedly and are asked to make shape judgments about them regardless of whether or not they actually perceive any 3-D shape. We wanted to access people’s lighting prior more directly by establishing the template they would employ to detect a shaded object in the absence of any visual cue to object shape. To this end, we adopted a classification image approach.