Face image can be decomposed into the lowfrequency component and highfrequency component by image decomposition method. Lowfrequency component characterizes face images illumination while highfrequency component describes its detail structure. Thus

highfrequency component is frequently used as illumination insensitive features for image recognition

but its performance will be affected by illumination variations because highfrequency component in the shadow area of the illuminated face image will be distorted after the decomposition. The method of restoring illuminated face image is proposed which consists of the following three steps: ① based on the PCA method

using the highfrequency information of local lighting area of an illuminated face image to restore highfrequency components of its asymmetrically illuminated area; ② using illumination of local lighting area to restore the holistic symmetrical frontal illumination components; and ③ fusing the restored lowfrequency illumination and highfrequency illumination

obtaining the symmetrical frontalilluminated face image without illumination effect and thus realizing face images lighting recovery. The proposed method is applied in the Yale B and acquires the restored illuminated face image with good performance in visual effect and recognition.