Font Size: 

Face recognition is a critical issue in image  processing. In this paper,  we propose a new method based on High Dimensional Model Representation (HDMR) philosophy for face recognition.   HDMR, as a decomposition method, has been also used to decompose images for making dimension reduction in recent years and it is obtained quite successful results. The proposed algorithm reads the matrix that describes the image as a multivariate data and uses HDMR as a data decomposition technique. HDMR decomposes a given image into low-variate data sets as constant, univariate, bivariate components.  Our algorithm then builds HDMR components and follows a set of steps to recognize people. 
To perform the efficiency of the proposed method some experiments are conducted on ORL database which contains grayscale face images of 40 people.  The database has 10 images for each person.  Besides, the method has been compared with the studies which are prepared by Tridiagonal Matrix Enhanced Multivariance Products Representation (TMEMPR) and Singular Value Decomposition (SVD) methods.