Abstract
Human vision can judge a lot of materials and surface qualities, such as glossiness, translucency, etc., just at a glance. However, it is still unclear whether human visual system use different mechanisms for evaluating different material classes. Several previous studies reported that perceived translucency is enhanced by adding clear specular highlights on the object surface. Also, a recent study reported that rating scores between perceived glossiness and translucency are significantly correlated. These findings suggest that some classes of material properties share a common underlying mechanism. We tested this hypothesis using the adaptation effect to the skewness of luminance histogram, which was previously used in a study on glossiness-perception. In our experiment, adaptation effects were tested for whether glossiness and translucency judgments would be affected in the same way or not using adapting stimuli with different polarities of skewness. They were texture images with the same mean luminance and luminance contrast but had skewed luminance histograms, either positively or negatively. The test images were of glossy or translucent objects generated on a computer by Blender software. Participants were instructed to rate perceived glossiness or translucency in a 9-point scale after adapting to images with positive/negative skewness or without adaptation. The apparent glossiness or translucency of the test images were perceptually equated in a preliminary session. The results showed a significant asymmetry in glossiness judgments between the two skewness conditions; glossiness scores were significantly reduced after adaptation to positive skewness than without adaptation, while the scores after adaptation to negative skewness showed no reduction. In contrast, translucency judgments showed almost no effects under the two skewness conditions. These results suggest that the mechanisms for the glossiness and translucency perception could be distinct. Specifically, the mechanisms associated with the luminance histogram skewness appear to differ, at least to some extent.