Image Statistics and the Fine Lines of Material Perception

Abstract

We experience vivid percepts of objects and materials despite complexities in the way images are structured by the interaction of light with surface properties (3D shape, albedo, and gloss or specularity). Although the perception of gloss (and lightness) has been argued to depend on image statistics (e.g., sub-band skew), studies have shown that perceived gloss depends critically on the structure of luminance variations in images. Here, we found that separately adapting observers to either positive or negative skew generated declines in perceived gloss, contrary to the predictions of theories involving image statistics. We also found similar declines in perceived gloss following adaptation to contours geometrically correlated with sharp specular edges. We further found this aftereffect was stronger when contour adaptors were aligned with specular edges compared with adaptation to the same contours rotated by 90°. These findings support the view that the perception of gloss depends critically on the visual system’s ability to encode specular edge structure and not image skew.