Image cross-covariance analysis for face recognition

In this paper, we proposed a novel technique for face recognition using image cross-covariance analysis (ICCA), based on the two-dimensional principal component analysis (2DPCA) technique. In conventional 2DPCA, the image covariance matrix is directly calculated via 2D images in matrix form, by concept of the covariance of a random variable. We found that it is not the optimal solution for 2DPCA framework. Because some useful information for classification is neglected. Thus, we introduced an image cross-covariance matrix which is a generalized form of the image covariance matrix. This matrix is defined by two variables. The first variable is the original image and the second one is the shifted version of the former. In this way, all information can be analyzed by 2DPCA frameworks. In this paper, the singular value decomposition (SVD) of image cross-covariance matrix is used to determine the optimal projection matrices. Experimental results on Yale, ORL and AR face databases show the improvement of our proposed techniques over the conventional 2DPCA technique.