Locality Preserving and Label-Aware Constraint-Based Hybrid Dictionary Learning for Image Classification

Dictionary learning has been an important role in the success of data representation. As a complete view representation, hybrid dictionary (HDL) is still its infant stage. In previous HDL approaches, scheme how to learn effective for image classification not well addressed. this paper, we proposed locality preserving and label-aware constraint-based (LPLC-HDL) method, apply it effectively. More specifically, information preserved by using graph Laplacian matrix based on shared commonality constraint with group regularization imposed coding coefficients corresponding class-specific particularity Moreover, all introduced constraints LPLC-HDL method are l2-norm regularization, which can be solved efficiently via employing alternative optimization strategy. The extensive experiments benchmark datasets demonstrate that our improvement over competing methods both hand-crafted deep features.