Face recognition based on improved projection dictionary pair learning algorithm
Daoju Deng, Xiumei Li · 2017
This paper presents an improved projection dictionary pair learning (DPL) algorithm, namely the C-DPL algorithm. The algorithm combines the fastness of DPL algorithm dictionary training and the discriminant strongness of collaborative representation classification (CRC) algorithm. Compared with DPL algorithm, C-DPL algorithm has achieved better recognition in face recognition. Experiments on the ORL datasets showed that the recognition rate of C-DPL is at least 0.7% higher than that of DPL at different training ratios; on the Yale datasets, the recognition rate of C-DPL is at least 1.01% higher than that of DPL at different training ratios; the validity and superiority of C-DPL algorithm are verified. On the Extended YaleB datasets, the recognition rate of C-DPL is 0.9% higher than that of DPL, indicating that C-DPL has good robustness to light change.