Improved Sparse Representation based on Robust Principal Component Analysis for Face Recognition

In this paper, we integrate Robust Principal Component Analysis (Robust PCA) and eigenface extraction into the sparse representation based classification. Firstly, the low-rank images are extracted by applying Robust PCA to make the training images as pure as possible. Then, Singular Value Decomposition (SVD) is adopted to extract the eigenfaces from the low-rank images. Finally, we combine these eigenfaces to construct a compact but discriminative dictionary for sparse representation. We evaluate our algorithm on several popular databases, experimental results demonstrate the effectiveness and robustness of our algorithm.