Image classification based on Laplacian PCA

Feature extraction plays a fundamental role in image classification and retrieval. However, the obtained feature space is often high-dimensional and dimensionality reduction is necessary to alleviate the curse of dimensionality or reduce the computational complexity. In this paper, we propose an image classification approach based on Laplacian PCA(LPCA). The notion of LPCA is borrowed from the area of manifold learning. Compared with the existing method, like PCA or KPCA, the proposed approach is more robustness against noise and weak metric-dependence on sample spaces. Experiments on three real image dataset with use of KNN as the classifier demonstrate the efficiency of the proposed method.