Lighting estimation and adjustment for facial images

Abstract

For robust face detection and recognition, the problem of lighting variation is considered as one of the greatest challenges. Lighting estimation and adjustment is a useful way to remove the influence of illumination for images. Due to the different prior knowledge provided by a single image and image sequences, algorithms dealing with lighting problems are always different for these two conditions. In this chapter we will present a lighting estimation algorithm for a single facial image and a lighting adjustment algorithm for image sequences. To estimate the lighting condition of a single facial image, a statistical model is proposed to reconstruct the lighting subspace where only one image of each subject is required. For lighting adjustment of image sequences, an entropy-based optimization algorithm is proposed to minimize the difference between consequent images. The effectiveness of those proposed algorithms are illustrated on face recognition, detection and tracking tasks.