关键词: 稀疏表示,人脸识别,字典学习,稀疏编码

Abstract: Using a compact dictionary obtained by sparse learning could greatly improve the accuracy and speed up the procedure of classification for sparse representation-based face recognition method.However,the traditional metaface learning (MFL) method doesn’t take into account the similarity among the training samples from the same person.In order to take the advantage of this prior information and make the learned dictionary more discriminative,an algorithm called coefficient-similarity-based metaface learning (CS-MFL) was proposed.In CS-MFL,the coefficient similarity is incorporated as a new constraint to the original objective function.To solve the new optimization problem,both l2 norm-based constraints are combined,and the original problem becomes a typical l₂-l1 problem.An experiment was carried out on different face databases,which shows that the proposed CS-MFL algorithm can achieve higher recognition rate than MFL algorithm,which demonstrates that the dictionary obtained by CS-MFL algorithm is more efficient and discriminative than that of the traditional MFL for face recognition application.

Key words: Sparse representation,Face recognition,Dictionary learning,Sparse coding