Color rendering of human faces under variable illumination

A system of color rendering of human faces under variable illumination is described. This system is categorized as an image-based rendering approach in which new images of a novel face are rendered using example images of faces (a bootstrap set) and the novel face. At least three images of each example face in the bootstrap set, taken under three linearly independent light source directions, are required in order to generate new images of faces under all possible lighting directions. The system is composed of three main components: color manipulation, illumination rendering, and color integration. The color manipulation component deals with color format to represent a color image of the novel face. It may also recover color reflectance of the novel face. After the color information of the novel face is manipulated, new images of the face under arbitrary illumination are generated by the illumination rendering component. The results from this component are then combined with necessary information to form color synthesized images. Based on the color manipulation component, three color rendering systems of human faces under variable illumination are given. These three proposed systems are explored, and effects of the bootstrap set on synthesized faces are investigated in this dissertation.