Abstract
When distinguishing illumination from reflectance edges, both blurriness of an edge and textural continuity across an edge are generally used as clues for interpreting the edge as an illumination edge. However, we found that, when these clues are combined, i.e., when a dark spot having a blurred edge is placed on textured backgrounds, the spot appears stained or painted rather than differently illuminated (Sawayama & Kimura, VSS2011). To elucidate the visual mechanism underlying this effect, this study analyzed spatial properties of a real shadow and stain, and investigated whether and how they are used for luminance edge interpretation. Wooden boards and cloths were used as textured backgrounds and onto these backgrounds a circular shadow was cast or a small amount of water was dropped to create shadow and stain stimuli. Then, photographs of the stimuli were taken under consistent lighting condition. The image analysis showed that edge blur of stains tends to be narrower in width than that of shadows. However, rating experiments revealed that the difference in spatial profile was not used when distinguishing stains from shadows. Observers judged the images of both shadows and stains as differing in reflectance. This bias in perception for reflectance over illumination edges (i.e., stains over shadows) was much reduced when a light spot was projected in place of a shadow even if their spatial profiles were similar except for the difference in luminance polarity. These results can be understood if we take into consideration much more frequent occurrence of a dark stain (and a light spot) than a "spot shadow" (an isolated shadow with the shadow caster being out of sight) in natural scenes. The bias for stains in perceiving dark spots may reflect heuristic processing that identifies the most likely cause of luminance variation based mainly on luminance polarity of the edge.
Meeting abstract presented at VSS 2012