Abstract
Perceived lightness of a surface in an image is influenced by the contexts of the surrounding images in addition to the surrounding luminance of perceptually adjacent area. One possible explanation for this would be a higher process of visual system recognizes the illuminant condition of the image and then affects the lightness perception. However, there could be another explanation that a lower process utilizes some statistics of the image’s luminance distribution having a relationship with the illuminant condition in the scene for deciding the perceived lightness directly. In this study, we investigated the relationship between the perceived lightness of a surface presented on natural images and the statistics such as the variance, skewness and maximum value of luminance distribution. In addition, we conducted a field measurement of illumimance to examine whether the illuminant condition in our environment relates to some statistics of the luminance distribution of the image. In this experiment, observers matched the perceived lightness of a test patch presented on the natural images to that of a comparison patch presented on a uniform gray background. The natural scene images were taken in indoor, outdoor and forest environments. We also used inverted and scrambled images to separate the influence of recognized image contents from that of luminance distribution. Results showed that there were correlations between the perceived lightness and the variance and skewness of luminance distribution of the images. The field measurement revealed a significant correlation between the perceived lightness and the illuminance measured in the forest environment, but no significant correlation between the measured illuminance and either the variance or the skewness of luminance distribution in all of the environments used. Theses results suggest that the visual system might refer some statistics of luminance distribution, however, they seem more complex than mere variance or skewness for lightness perception.
Meeting abstract presented at VSS 2012