Video Face Recognition on Invariable Moment

There are three limitations in video face recognition. Firstly, face does not lie in image center, secondly, the size of face is bigger or smaller than those in the facial database, and lastly it exists possibly rotation in the face. In order to overcome these drawback caused by the above factors, the paper presents a novel method based on invariable moment that chooses candidate faces using principle. If number of candidate faces is more than one, we reconstruct face that has not been recognized using invariable moment and compare it with candidate faces to accomplish recognition. The experimentations show that the method can solve the insufficiencies in the certain extent and suppress refusal recognition in the video face recognition.