Semi-supervised Sparsity Preserving Two-Dimensional Marginal Fisher Analysis Dimensionality Reduction Algorithm

For the problem of labeled samples scarcity among samples sets,a semi-supervised sparsity preserving two-dimensional marginal fisher analysis algorithm is proposed in this paper.First, dimension is reduced based on image matrix,which makes effectively use of image pixels spatial structure information.Then,it designs intra-class scatter matrix and inter-class scatter matrix to preserve the intra-class compactness and the inter-class separability for training samples.Finally,it constrains the sparse reconstruction among features by sparsity preserving,which not only preserve local geometric structure,but also contain natural discriminative information.Experimental results on YALE,ORL and AR face databases demonstrate that the proposed algorithm has good classification and recognition performance.