Abstract:Recently Sparse Representation (or coding) based Classification (SRC) has gained great success in face recognition. In SRC, the testing image is expected to be best represented as a sparse linear combination of training images from the same class, and the representation fidelity is measured by the ℓ 2-norm or ℓ 1-norm of the coding residual. However, SRC emphasizes the sparsity too much and overlooks the spatial information during local feature encoding process which has been demonstrated to be critical in real-world face recognition problems. Besides, some work considers the spatial information but overlooks the different discriminative ability in different face regions. In this paper, we propose to weight spatial locations based on their discriminative abilities in sparse coding for robust face recognition. Specifically, we learn the weights at face locations according to the information entropy in each face region, so as to highlight locations in face images that are important for classification. Furthermore, in order to construct a robust weights to fully exploit structure information of each face region, we employed external data to learn the weights, which can cover all possible face image variants of different persons, so the robustness of obtained weights can be guaranteed. Finally, we consider the group structure of training images (i.e. those from the same subject) and added an ℓ 2,1-norm (group Lasso) constraint upon the formulation, which enforcing the sparsity at the group level. Extensive experiments on three benchmark face datasets demonstrate that our proposed method is much more robust and effective than baseline methods in dealing with face occlusion, corruption, lighting and expression changes, etc.

L1-norm plus L2-norm sparse parameter for image recognition

Two-stage nonnegative sparse representation for large-scale face recognition.

Superimposed Sparse Parameter Classifiers for Face Recognition

Fisher discrimination-based l_2,1 -norm sparse representation for face recognition

Robust Face Recognition Based l21-Norm Sparse Representation

Fisher Discrimination-Based $$l_{2,1} $$L2,1-Norm Sparse Representation for Face Recognition

Fast Low Rank Representation Based Spatial Pyramid Matching for Image Classification.

Loose L 1 / 2 Regularised Sparse Representation for Face Recognition

Sparsity Analysis Versus Sparse Representation Classifier.

Collaborative Representation for Classification, Sparse or Non-sparse?

Collaborative neighbor representation based classification using l2-minimization approach

Discriminative Sparse Representation for Face Recognition

A New Discriminative Sparse Representation Method for Robust Face Recognition Via $l_{2}$ Regularization

Hyperspectral Image Classification with Spatial Filtering and l(2, 1) Norm.

Beyond sparsity: The role of L1-optimizer in pattern classification

Double linear regression classification for face recognition

Stable and compact face recognition via unlabeled data driven sparse representation-based classification

Multi-linear Regression Coefficient Classifier for Recognition.

Unlabeled data driven cost-sensitive inverse projection sparse representation-based classification with 1/2 regularization

Hyperspectral Image Classification with Spatial Filtering and \(l_{(2,1)}\) Norm

Face Recognition by Sparse Discriminant Analysis Via Joint L2,1-norm Minimization