Multi-Resolution Dictionary Learning Algorithm with Discriminative Locality Constraints for Face Recognition

Shuying Zeng; Hongzhong Tang; Shijun Deng; Dongbo Zhang

doi:10.3788/LOP202158.1415008

Laser & Optoelectronics Progress, Volume. 58, Issue 14, 1415008(2021)

Multi-Resolution Dictionary Learning Algorithm with Discriminative Locality Constraints for Face Recognition

Shuying Zeng¹, Hongzhong Tang^1,2、*, Shijun Deng², and Dongbo Zhang²

Author Affiliations

¹College of Automation and Electronic Information, Xiangtan University, Xiangtan, Hunan 411105, China

²Hunan Provincial Key Laboratory of Intelligent Information Processing and Application, Hengyang, Hunan 421002, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(13)

Fig. 1. MDLDLC algorithm flow chart

Download full size

Fig. 2. Correspondence between dictionary atom and profile vector

Download full size

Fig. 3. Comparative analysis of LLC, LCLE-DL and MDLDLC algorithms. (a) LLC algorithm; (b) LCLE-DL algorithm; (c) proposed MDLDLC algorithm

Download full size

Fig. 4. Face images with different resolutions from the ORL database. (a) 56 pixel×46 pixel; (b) 28 pixel×23 pixel; (c) 14 pixel×12 pixel

Download full size

Fig. 5. Face images with different resolutions from the Extended Yale B database. (a) 64 pixel×64 pixel; (b) 32 pixel×32 pixel; (c) 16 pixel×16 pixel

Download full size

Fig. 6. Face images with different resolutions from the AR database. (a) 50 pixel×40 pixel; (b) 25 pixel×20 pixel; (c) 12 pixel×10 pixel

Download full size

Fig. 7. Face images with different resolutions from PIE database. (a) 64 pixel×64 pixel; (b) 32 pixel×32 pixel; (c) 16 pixel×16 pixel

Download full size

Fig. 8. Effect of parameter selection of α and β on the different datasets. (a) ORL database; (b) Extended Yale B database; (c) AR database; (d) PIE database

Download full size

Fig. 9. Effect of number of atoms on the different datasets. (a) ORL database; (b) Extended Yale B database; (c) AR database; (d) PIE database

Download full size

Table 1. Recognition rates of different algorithms on the ORL database unit: %
View table
Table 1. Recognition rates of different algorithms on the ORL database unit: %
Algorithm Recognition rate
D-KSVD 86.10±2.63
LCKSVD1 86.52±1.73
LCKSVD2 87.10±2.23
LCLE-DL 90.42±1.56
SLatDPL 90.80±1.04
MRDL 91.75±1.25
ResNet 84.00±1.24
MDLDLC 94.62±0.81

Table 2. Recognition rates of different algorithms on the Extended Yale B database unit: %
View table
Table 2. Recognition rates of different algorithms on the Extended Yale B database unit: %
Algorithm Recognition rate
D-KSVD 72.54±1.38
LCKSVD1 73.13±1.24
LCKSVD2 75.50±1.36
LCLE-DL 70.87±1.27
SLatDPL 86.70±1.26
MRDL 87.86±1.43
ResNet 86.48±1.25
MDLDLC 91.60±0.90

Table 3. Recognition rates of different algorithms on the AR database unit: %
View table
Table 3. Recognition rates of different algorithms on the AR database unit: %
Algorithm Recognition rate
D-KSVD 66.17±1.23
LCKSVD1 74.73±1.25
LCKSVD2 75.60±1.30
LCLE-DL 71.21±1.23
SLatDPL 84.47±1.45
MRDL 82.36±1.52
ResNet 85.63±1.23
MDLDLC 85.39±1.12

Table 4. Recognition rates of different algorithms on the PIE database unit: %
View table
Table 4. Recognition rates of different algorithms on the PIE database unit: %
Algorithm Recognition rate
D-KSVD 62.31±0.98
LCKSVD1 65.72±0.90
LCKSVD2 66.20±0.75
LCLE-DL 84.22±0.56
SLatDPL 91.57±0.58
MRDL 95.49±0.45
ResNet 69.00±0.74
MDLDLC 97.16±0.36

Tools

Get Citation

Copy Citation Text

Shuying Zeng, Hongzhong Tang, Shijun Deng, Dongbo Zhang. Multi-Resolution Dictionary Learning Algorithm with Discriminative Locality Constraints for Face Recognition[J]. Laser & Optoelectronics Progress, 2021, 58(14): 1415008

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites