Create an image gallery! In this way, due to the similarity of facial structures, once normalized after the facial image, you can plainly see, the facial image is not random and scattered distribution in a higher dimensional space, but the existence of a law. In the next experiment, n is 128 Us know, k-l transform method in essence, is a facial image pixels in gray-scale spatial correlation of time for you to decide. Therefore, prior to the k-l transformation, us need to face a series of treatment in order to achieve the purpose of calibration and gray normalized for location. Then us first look at the image of normalization, behind us image applied to the normalized k-l transform.