Face color constancy: Towards illumination invariant face recognition

We propose a novel technique to compensate for variations in face images due to changes in the illumination conditions. Other than color of the faces, shadows and specular reflections on the faces also change with respect to lighting environment. These variations result in false recognitions even in the best face recognition algorithms. We tackle this problem by processing the faces using a color constancy model specific to face images, which we named face color constancy. The face color constancy algorithm is trained with face images in different lighting environments. The color changes observed are used to learn the color mapping from one lighting environment to the other. These mappings are represented in a low dimensional subspace to obtain basis vector fields. Using these basis vector fields we can model the nonlinear color changes to transform a new probe face to a reference face in the gallery. Using face color constancy preprocessing, we show that for fixed pose and expression we require a single training image per subject to perform accurate ace recognition.