Face Classification Using Sparse Reconstruction Embedding
Min Liang, Yu Zhuang, Cheng En Cai · 2010
Face classification is an active and important research area in image processing, pattern classification and computer vision, since it is widely used in real-world application. In general, face classification framework consists of feature extraction and classifier, while feature extraction method is considered critical to the performance of the classifier. Traditional face classification frameworks used to employ the classical methods such as Principle Component Analysis (PCA) and Laplacian Eigenmap for feature extraction. In order to achieve better performance of face classification, we propose a new framework based on Sparse Reconstruction Embedding (SRE) method. In our framework, we firstly use maximum likelihood estimator(MLE) to estimate the optimal dimensionality of the data, and then use SRE to obtain the low-dimensional representations, finally, Quadratic discriminant classifier (QDC) is employed to compute the classification model. We conduct experiments on publicly available database to examine the efficacy of the proposed framework.