Multiple collaborative representations for face recognition
Zhongli Ma, Quanyong Liu, Liang Hao · 2016
Different pixel plays different roles in representing a face image. In this paper, we proposed a novel representation which integrates original and its virtual face image to represent test sample. This method first combines two adjacent columns of an original face image to generate corresponding virtual face image, and then classification algorithm is respectively applied to the original and virtual face images to obtain two different representation results. At last, this method directly integrates the two results to get the final representation result and uses the final result to classify test sample. This method can enlarge the number of training samples for each subject and adequately exploit the detail features of each target image so as to improve the recognition accuracy. The abundant experiments of face recognition show that the new representation method can obtain a higher accuracy than other algorithms.