Laser & Optoelectronics Progress, Volume. 57, Issue 22, 221011(2020)
Super-Resolution Reconstruction of Densely Connected Generative Adversarial Network Images
Figures & Tables(13)
![Comparison of butterfly reconstruction effect. (a) Original image; (b)method in Ref. [23]; (c) method in Ref. [5]; (d) method in Ref. [7]; (e) method in Ref. [8]; (f) method in Ref. [13]; (g) method in Ref. [16]; (h)proposed method](/Images/icon/loading.gif){kind=icon}
Fig. 5. Comparison of butterfly reconstruction effect. (a) Original image; (b)method in Ref. [23]; (c) method in Ref. [5]; (d) method in Ref. [7]; (e) method in Ref. [8]; (f) method in Ref. [13]; (g) method in Ref. [16]; (h)proposed method
Fig. 6. Comparison of lenna reconstruction effect. (a) Original image; (b)method in Ref. [23]; (c) method in Ref. [5]; (d) method in Ref. [7]; (e) method in Ref. [8]; (f) method in Ref. [13]; (g) method in Ref. [16]; (h)proposed method
Fig. 7. Comparison of 253027 reconstruction effect. (a) Original image; (b)method in Ref. [23]; (c) method in Ref. [5]; (d) method in Ref. [7]; (e) method in Ref. [8]; (f) method in Ref. [13]; (g) method in Ref. [16]; (h)proposed method
Fig. 8. Comparison of barbara reconstruction effect. (a) Original image; (b)method in Ref. [23]; (c) method in Ref. [5]; (d) method in Ref. [7]; (e) method in Ref. [8]; (f) method in Ref. [13]; (g) method in Ref. [16]; (h)proposed method
Fig. 10. Comparison of image edge extraction before and after convolution operation. (a)--(c) Images after convolution operation; (d)--(f) corresponding images before convolution operation
Table 1. Comparison of PSNR between proposed algorithm and mainstream algorithm on four test sets unit:dB
View tableTable 1. Comparison of PSNR between proposed algorithm and mainstream algorithm on four test sets unit:dB
Dataset Scale Method in Ref. [23] Method in Ref. [5] Method in Ref. [7] Method in Ref.[8] Method in Ref.[13] Method in Ref. [16] Proposed method Set5 2 3 4 33.57 30.42 29.00 31.39 36.61 31.99 30.27 36.72 32.85 30.56 37.44 33.45 31.12 37.71 33.90 31.62 37.93 34.11 31.79 Set14 2 3 4 30.24 27.55 26.00 28.31 32.28 29.13 27.32 32.45 29.30 27.50 33.03 29.77 28.01 33.23 29.92 28.14 33.44 30.08 28.35 B100 2 3 4 29.49 27.11 25.88 27.83 30.97 28.10 26.79 31.45 28.38 26.96 31.57 28.96 27.22 31.90 29.01 27.38 32.13 29.30 27.57 Urban100 2 3 4 26.76 24.59 23.30 25.35 28.94 26.01 24.45 29.49 26.33 24.65 30.69 27.23 25.09 30.85 27.44 25.36 30.97 27.58 25.63
Table 2. Comparison of SSIM between proposed algorithm and mainstream algorithm on four test sets
View tableTable 2. Comparison of SSIM between proposed algorithm and mainstream algorithm on four test sets
Dataset Scale Method in Ref. [23] Method in Ref. [5] Method in Ref. [7] Method in Ref.[8] Method in Ref.[13] Method in Ref. [16] Proposed method Set5 2 3 4 0.9293 0.8671 0.8115 0.8819 0.9500 0.9003 0.8622 0.9539 0.9088 0.8630 0.9576 0.9209 0.8839 0.9577 0.9228 0.8871 0.9603 0.9239 0.8877 Set14 2 3 4 0.8679 0.7738 0.7026 0.7937 0.9070 0.8177 0.7487 0.9057 0.8209 0.7521 0.9130 0.8325 0.7669 0.9131 0.8329 0.7678 0.9143 0.8326 0.7687 B100 2 3 4 0.8429 0.7383 0.6680 0.7458 0.8854 0.7844 0.7069 0.8880 0.7857 0.7100 0.8955 0.7968 0.7247 0.8958 0.7981 0.7266 0.8993 0.7994 0.7280 Urban100 2 3 4 0.8411 0.7338 0.6567 0.7554 0.8922 0.7908 0.7183 0.8951 0.7976 0.7233 0.9135 0.8266 0.7509 0.9150 0.8246 0.7537 0.9164 0.8299 0.7551
Table 3. Comparison of time consumption between proposed algorithm and mainstream algorithms on four test sets unit:s
View tableTable 3. Comparison of time consumption between proposed algorithm and mainstream algorithms on four test sets unit:s
Dataset Scale Method in Ref. [23] Method in Ref. [5] Method in Ref. [7] Method in Ref.[8] Method in Ref.[13] Method in Ref. [16] Proposed method Set5 2 3 4 - - - 0.47 0.58 0.32 0.24 2.19 2.23 2.19 0.13 0.14 0.12 0.21 0.19 0.22 0.11 0.13 0.12 Set14 2 3 4 - - - 0.51 0.84 0.56 0.38 4.32 4.40 4.39 0.25 0.26 0.25 0.27 0.23 0.21 0.22 0.21 0.20 B100 2 3 4 - - - 0.52 0.59 0.33 0.26 2.51 2.58 2.51 0.16 0.21 0.21 0.30 0.27 0.25 0.13 0.17 0.15 Urban100 2 3 4 - - - 0.69 2.96 1.67 1.21 22.12 19.35 18.46 0.98 1.08 1.06 1.01 1.00 1.03 0.91 0.98 1.02
Tools
Get Citation
Copy Citation Text
Bin Li, Lu Ma. Super-Resolution Reconstruction of Densely Connected Generative Adversarial Network Images[J]. Laser & Optoelectronics Progress, 2020, 57(22): 221011
Paper Information
Category: Image Processing
Received: Feb. 14, 2020
Accepted: Mar. 26, 2020
Published Online: Nov. 3, 2020
The Author Email: Bin Li (747952996@qq.com)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20