Interest in the visual perception of the materials that objects are made of has been growing. Most past studies on visual material-category perception have used stimuli with rich information such as color, shape, and texture. However, we can distinguish material categories from even simple black-and-white line drawings. This paper presents a new attempt to analyze material perception from Japanese “manga” comics which are composed of line drawings typically printed in black and white. In this study, we first collected 400 material objects captured from manga comics, and created 400 corresponding patches that were close-up images of the objects and that excluded shape information. Through psychophysical experiments, 274 pairs of images giving consistent material impressions to observers were chosen. According to our experiments, observers could distinguish material categories from patches with on accuracy of 88.4%. Then for each material, we investigated the low-dimensional image features that contributed to the material perception of black-and-white drawings. In particular, we found image features that represented metals very well.
Yuma Saito, Keita Hirai, Takahiko Horiuchi, "Construction of Manga Materials Database for Analyzing Perception of Materials in Line Drawings" in Proc. IS&T 23rd Color and Imaging Conf., 2015, pp 201 - 206, https://doi.org/10.2352/CIC.2015.23.1.art00036