An illumination compensation algorithm for face images based on line scanning

The difference among faces under different illumination is a bottleneck in face recognition. This paper presents an illumination compensation algorithm based on two-dimensional image information for human faces. On the assumption that the human face shares a similar shape with spherical surface, the algorithm mainly consists of two illumination estimation processes and one illumination compensation process. We first estimate the face information under even illumination along the symmetrical axis of faces so as to build a standard illumination model by data fitting to the prior statistic information. Then, we analyze the statistical distributions of the face image grayscale along the direction of lighting. At last, using the standard model combined with linear transform and non-linear transform, we can rectify the face image under uneven illumination to standard. The simulation results on the Yale B face database have shown that the proposed method can realize effectively illumination compensation under wide-angle oblique lighting and very dark lighting. In addition to simple computation, the algorithm surpasses other methods in visual effect and information correlation.