Abstract
When distinguishing illumination from reflectance edges, luminance gradient (or fuzziness) of the edge is used as a clue. Generally, a fuzzy edge tends to be perceived as an illumination edge. However, we found that when a fuzzy-edged region was placed on textured backgrounds it tended to appear stained or painted rather than differently illuminated. In our experiment, a dark center region, which simulated a cast shadow in terms of luminance contrast, was placed on spatially uniform or textured backgrounds. The luminance edge between the center and surrounding regions was either fuzzy or sharp. Observers were asked to rate how certainly the dark region was perceived as having lower reflectance or dimly illuminated using a 7-point scale. Results showed that the dark region appeared stained on the textured background, whereas it appeared more like a shadow on the uniform background. It should be noted that the results on the textured background go against the belief that both fuzzy edges and textural continuity over the center and surrounding regions favor illumination-edge interpretation. We tested several different textures such as tatami-mat, small pebbles, grass, and marble patterns, and obtained similar results. However, coarse checkerboard patterns produced the results similar to those on the spatially uniform field. Subsequent experiments revealed that spatial frequency of the checkerboard pattern is a critical factor. The fuzzy-edged dark region was perceived as a shadow on checkerboard patterns of lower and higher spatial frequencies and it appeared more like a stain when the spatial frequency was within a limited range around 2 c/deg. Taken together, these results suggest that spatial interaction between texture and luminance gradient of the edge plays an important role in distinguishing reflectance from illumination edges. This interaction seems to change the way to decompose image luminance into illumination and surface reflectance.