KCRC-LCD: Discriminative kernel collaborative representation with locality constrained dictionary for visual categorization

Highlights：

• Propose a generalized KCRC-LCD framework with good performance and scalability.

• LCD can be nicely incorporated under the framework of KCRC through kernelization.

• A unified similarity measurement framework is considered to reduce metric biases.

• Conducted comprehensive experiments and analysis.

• The proposed framework is comparable or outperforms state-of-the-art methods.

摘要

Highlights•Propose a generalized KCRC-LCD framework with good performance and scalability.•LCD can be nicely incorporated under the framework of KCRC through kernelization.•A unified similarity measurement framework is considered to reduce metric biases.•Conducted comprehensive experiments and analysis.•The proposed framework is comparable or outperforms state-of-the-art methods.

论文关键词：CRC,Collaborative representation classification,KCRC,Kernel collaborative representation classification,LCD,Locality constrained dictionary,GD,Global dictionary,KCRC-LCD,CRC with locality constrained dictionary,KCRC-GD,CRC with global dictionary,KCRC-LCD,KCRC with locality constrained dictionary,KCRC-GD,KCRC with global dictionary,SRC,Sparse representation-based classification,KSRC,Kernel sparse representation-based classification,SVM,Support vector machine,KPCA,Kernel principle component analysis,KFDA,Kernel fisher discriminant analysis,KCRC-Identity,KCRC with no dimensionality reduction,KCRC-KPCA,KCRC with KPCA dimensionality reduction,KCRC-RP,KCRC with random projection,KCRC-Graph,KCRC with graph projection,CRC-RLS,CRC with regularized least square,KCRC-RLS,Kernel CRC with regularized least square,RCRC,Robust CRC,LC-KSVD,Locality constrained K-SVD,D-KSVD,Discriminative K-SVD,Kernel collaborative representation,Regularized least square algorithm,Nearest neighbor,Locality constrained dictionary