Dual Unsupervised Discriminant Projection for Face Recognition
In this paper we proposed a dual unsupervised discriminant projection (DUDP) method for dimensionality reduction tasks. The proposed method is derived from the efficient unsupervised method called unsupervised discriminant projection (UDP). UDP takes into account both the local and nonlocal characteristics to seek a projection that simultaneously maximizes the nonlocal scatter and minimizes the local scatter. While UDP adopt PCA procedure to avoid a singular scatter matrix by ruling out some small principal components in which it lost a lot of potential and valuable discriminant information of original data. To overcome this problem, we proposed our algorithm to carry out discriminant analysis both in null space and range space to avoid loss of discriminant information. The advantage of our algorithm is borne out by comparison with some other widely used methods in the experiments on Yale face database.
KeywordsDimensionality reduction unsupervised discriminant projection subspace learning face recognition
Unable to display preview. Download preview PDF.
- 6.Belkin, M., Niyogi, P.: Laplacian Eigenmaps and Spectral Techniques for Embedding and Clustering. Advances in Neural Information Processing Systems 14, 585–591 (2001)Google Scholar
- 7.Bengio, Y., Paiement, J.F., Vincent, P.: Out-of-sample Extensions for LLE, Isomap, MDS, Eigenmaps, and Spectral Clustering. Technical Report 1238 Université de Montreal (2003)Google Scholar
- 11.Yale Univ. Face Database, http://cvc.yale.edu/projects/yalefaces/yalefaces.html