The Optimal Graph Regularized Sparse Coding with Application to Image Representation

摘要

Sparse representation has shown its superiority and effectiveness in many real applications in recent years. However, it is still an open problem to effectively preserve the intrinsic geometric structure of data in new representation space. In this paper, we propose a novel method, called the Optimal Graph regularized Sparse Coding (OGSC), to deal with the high dimensional data. Specifically, we impose a rank constraint on the Laplacian matrix of the graph model, and thus can learn the optimal graph to preserve the manifold structure of data in each iteration. Additionally, the optimization scheme for our proposed method is also provided in this paper. The experimental results on three benchmark datasets have shown that our proposed OGSC method outperforms other stat-of-the-art methods.