Color naming experiments using 2D and 3D rendered samples

This article describes a color naming experiment using 2D and 3D rendered color samples. Conventional color naming experiments using a priori clues generally involve 2D clues such as color patches. However, in real‐world scenes, most objects have 3D shapes whose colors are affected by illumination effects such as shadows and gloss. We use 2D and 3D rendered samples as clues in the experiments, and analyze the relationship between color terms and object surfaces. First, we develop a color term collection system that can produce 218 test colors. We render the color images of a flat disk as a 2D sample and a sphere as a 3D sample on a calibrated display device. It is supposed that the 2D and 3D surfaces with the same object color are obtained under the same conditions of viewing and illumination. The results of color naming experiments show that there are differences for color terms between 2D and 3D samples. Important findings are as follows: (1) brighter color terms tend to be chosen for the 3D samples than the 2D samples, when observing achromatic colors, (2) achromatic color terms are chosen for 3D samples having low saturation, and (3) for chromatic colors, a darker color term is generally chosen in comparison to the corresponding 2D samples of the same color. These properties become more prominent by changing the illumination angle from 0° to 45° to the surface normal. © 2014 Wiley Periodicals, Inc. Col Res Appl, 40, 270–280, 2015