Abstract
Skin colors of NPR (Non-Photorealistic Rendering) images, such as illustrations, animated cartoons, and toon shading images, have appear to have different features from the photography and real skin colors. For developing an automatic coloring system based on a color order system for NPR images, the tolerable range of skin colors needs to be determined based on characters and backgrounds. In this study, we investigated tolerable skin colors of the facial shape with or without background images were researched.
Observers judged the tolerable colors as animation facial skin colors. Skin colors were estimated, firstly, without background in normal environments, containing various conditions, such as tanning, gender and ages, and, secondly, with background assuming the character was in the background environment. We used background images adjusted with color temperature and lightness. The colors of stimuli were expressed in CIE L*a*b* color space.
In the a*-b* coordinates, tolerable skin colors, measured with various backgrounds, were found to be almost contained in the range of tolerable skin color without backgrounds. It appeared that the range of tolerable skin color was stable in the a*-b* coordinates. However, in the a*-L* coordinates, the tolerable skin color showed different features with backgrounds from without background. Dark backgrounds had significant influence on the range of lightness. As results, the tolerable lightness of the skin color was changed as the background changed.