Improved Discriminative Sparsity Preserving Projections for Image Recognition

In this paper, we improve discriminative sparsity preserving projections (DSPP) which integrates global within-class structure into the discriminant manifold learning objective function for dimensionality reduction, and propose a new sparse subspace learning algorithm called improved discriminative sparsity preserving projections (IDSPP). Different from DSPP, IDSPP introduces between-class structure into the discriminant manifold learning objective function for dimensionality reduction. Extensive experiment on CMU-PIE face image dataset demonstrates the effectiveness of the proposed method.