One fast and automatic face recognition method

Although independent component analysis(ICA) can reflect higher order statistical feature of image data, it is found that in the directly application of ICA in face recognition, even using fast ICA algorithm(FastICA), there also exist problems such as large computation and time consuming. Therefore in this paper one fast and automatic face recognition method was proposed, which first reduce the dimension of original face image by kernel principle component analysis(KPCA), thus to give prominence to the principle component feature of face image and also take account of higher order statistical feature which including the nonlinear relationship among image pixels, then use algorithm of FastICA to extract the principle independent components of facial features which are more useful for and will be used in face classification at last. Tests results on a subset of FERET database show the validity of the proposed method.