Abstract

Whether choosing fresh fish, or walking on wet tiles, the visual impression of surface properties influences diverse behaviours. Specularity—the extent to which a surface reflects light like a mirror—conveys important information about an object’s physical properties such as its composition, smoothness and physical state (e.g. wet or dry). To estimate a surface’s reflectance or gloss, the brain must make an inference based on retinal images that result from unknown physical properties of the environment related to (i) three-dimensional shape; (ii) surface reflectance and (iii) the illumination field. While disentangling these unknowns is formally intractable, a biological and/or computational solution is likely to be found in the characteristic image properties obtained when viewing reflective objects. This idea has a long history (e.g., Helmholtz), however we have limited formal understanding of which signals the human visual system uses to estimate gloss. Here I use computational image analysis and human psychophysics to suggest a role for disparity statistics in driving gloss perception. Identifying these cues suggests a straightforward means for the brain to estimate surface attributes using low-level image signals rather than depending on complex physical models.