Low-rank 2D local discriminant graph embedding for robust image feature extraction

Highlights：

• The main contributions of the paper are summarized as follows.

• It is learned a low-rank matrix based on the graph embedding that can simultaneously perform subspace learning, graph Laplacian regularization, and low-rank learning in a unified strategy is proposed and an iterative solution to the convex optimization problem is provided.

• It is combined the graph embedding framework with the low-rank matrix, two intraclass and interclass weighted matrix graphs are proposed, which fully discover the manifold structural information of the neighbourhood and improve the recognition ability in 2D images.

• It is proposed to ensure that the given data are divided into a low-rank feature coding part and a sparse noise error part to improve the recognition ability, which can weaken the influence of noise and occlusion when learning the optimal projection.