Chinese Optics, Volume. 15, Issue 5, 954(2022)
Survey of non-blind image restoration
Figures & Tables(13)
![Learned dictionary from non-blind image restoration algorithm[68]. (a) Partial restoration image for Barbara image; (b) the learned dictionary](/Images/icon/loading.gif){kind=icon}
Fig. 4. Learned dictionary from non-blind image restoration algorithm[68]. (a) Partial restoration image for Barbara image; (b) the learned dictionary
Fig. 7. The image restoration framework based on CV-CNN network[97]
Table 1. Experimental settings
View tableView in ArticleTable 1. Experimental settings
序号 点扩散函数 噪声水平 图像 1 9 × 9 boxcar BSNR = 40 dB Cameraman 2 $k(x,y) = 1/({x^2} + {y^2}),x,y = - 7,\cdots,7$ $ {\sigma ^2} = 2 $ Cameraman 3 $k(x,y) = 1/({x^2} + {y^2}),x,y = - 7,\cdots,7$ $ {\sigma ^2} = 8 $ Cameraman 4 $k = {[1,4,6,4,1]^{\rm{T}}}[1,4,6,4,1]/256$ $ {\sigma ^2} = 49 $ Lena 5 Gaussian型点扩散函数，方差为1.6 $ {\sigma ^2} = 2 $ Barbara 6 Gaussian型点扩散函数，方差为0.4 $ {\sigma ^2} = 64 $ House
Table 2. Comparison of ISNR output by eight methods
View tableView in ArticleTable 2. Comparison of ISNR output by eight methods
实验 方法 1 2 3 4 5 6 ForWaRD[ 15 ]7.40 6.75 5.07 2.98 0.98 5.52 ShearDec[ 21 ]7.89 7.55 5.56 − − − GSM[ 23 ]−1.61 6.84 5.29 − 0.95 5.98 SV-GSM[ 24 ]7.33 7.45 5.55 − 1.36 6.02 LPA-ICI[ 26 ]8.29 7.82 5.98 3.90 − − SA-DCT[ 27 ]8.55 8.11 6.33 4.49 1.02 5.96 SURE-LET[ 25 ]7.84 7.54 5.22 4.42 1.06 4.38 BM3DDEB[ 31 ]8.34 8.19 6.40 4.81 1.28 7.21
Table 3. Experimental setup for iterative methods
View tableView in ArticleTable 3. Experimental setup for iterative methods
序号 点扩散函数 噪声水平 1 $ k(x,y) = 1/({x^2} + {y^2}),x,y = - 7,\cdots,7 $ ${\sigma ^2} = 2$ 2 $ k(x,y) = 1/({x^2} + {y^2}),x,y = - 7,\cdots,7 $ ${\sigma ^2} = 8 $ 3 9 × 9 boxcar BSNR = 40 dB 4 $ k = {[1,4,6,4,1]^{\rm{T} } }[1,4,6,4,1]/256 $ ${\sigma ^2} = 49 $ 5 Gaussian型点扩散函数，方差为1.6 ${\sigma ^2} = 2 $ 6 Gaussian型点扩散函数，方差为0.4 ${\sigma ^2} = 64 $
Table 4. [in Chinese]
View tableView in ArticleTable 4. [in Chinese]
实验序号 1 2 3 4 5 6 方法 Cameraman BM3DDEB[ 31 ]8.19 6.40 8.34 3.34 3.73 4.70 L0-Abs[ 62 ]7.70 5.55 9.10 2.93 3.49 1.77 CGMK[ 36 ]7.80 5.49 9.15 2.80 3.54 3.33 TVMM[ 34 ]7.41 5.17 8.54 2.57 3.36 1.30 GFD[ 33 ]8.38 6.52 9.73 3.57 4.02 - NCSR[ 70 ]8.78 6.69 10.33 3.78 4.60 4.50 GSR[ 71 ]8.39 6.39 10.08 3.33 3.94 4.76 IDDBM3D[ 73 ]8.85 7.12 10.45 3.98 4.31 4.89 LRD[ 76 ]8.90 7.05 10.70 3.99 4.62 4.62 House BM3DDEB[ 31 ]9.32 8.14 10.85 5.13 4.56 7.21 L0-Abs[ 62 ]8.40 7.12 11.06 4.55 4.80 2.15 CGMK[ 36 ]8.31 6.97 10.75 4.48 4.97 4.59 TVMM[ 34 ]7.98 6.57 10.39 4.12 4.54 2.44 GFD[ 33 ]9.39 7.75 12.02 5.21 5.39 NCSR[ 70 ]9.96 8.48 13.12 5.81 5.67 6.94 GSR[ 71 ]10.02 8.56 13.44 6.00 5.95 7.18 IDDBM3D[ 73 ]9.95 8.55 12.89 5.79 5.74 7.13 LRD[ 76 ]10.09 8.67 13.49 6.03 6.22 6.74 Lena BM3DDEB[ 31 ]7.95 6.53 7.97 4.81 4.37 6.40 L0-Abs[ 62 ]6.66 5.71 7.79 4.09 4.22 1.93 CGMK[ 36 ]6.76 5.37 7.86 3.49 3.93 4.46 TVMM[ 34 ]6.36 4.98 7.47 3.52 3.61 2.79 GFD[ 33 ]8.12 6.65 8.97 4.77 4.95 - NCSR[ 70 ]8.03 6.54 9.25 4.93 4.86 6.19 GSR[ 71 ]8.24 6.76 9.43 5.17 4.96 6.57 IDDBM3D[ 73 ]7.97 6.61 8.91 4.97 4.85 6.34 LRD[ 76 ]8.25 6.78 9.31 5.13 5.08 6.13 Barbara BM3DDEB[ 31 ]7.80 3.94 5.86 1.90 1.28 5.80 L0-Abs[ 62 ]3.51 1.53 3.98 0.73 0.81 1.17 CGMK[ 36 ]2.45 1.34 3.55 0.44 0.81 0.38 TVMM[ 34 ]3.10 1.33 3.49 0.41 0.75 0.59 NCSR 7.76 3.64 5.92 2.06 1.43 5.50 GSR[ 71 ]8.98 4.80 7.15 2.19 1.58 6.20 IDDBM3D[ 73 ]7.64 3.96 6.05 1.88 1.16 5.45 LRD[ 76 ]8.31 5.17 6.95 2.34 1.70 5.37
Table 5. Experimental comparison of deep learning of different methods
View tableView in ArticleTable 5. Experimental comparison of deep learning of different methods
Levin[ 106 ]Sun[ 107 ]Martin[ 108 ]σ 1% 3% 5% 1% 5% 1% 5% EPLL[ 82 ]34.06 29.09 26.54 32.48 26.78 29.81 24.66 0.9310 0.8460 0.7785 0.8815 0.6975 0.8383 0.6276 CSF[ 84 ]31.09 28.01 26.32 31.52 26.62 29.00 24.93 0.9024 0.8013 0.7427 0.8622 0.6735 0.8230 0.6428 MLP[ 89 ]32.08 27.00 25.38 31.47 24.65 28.47 24.01 0.8884 0.7016 0.6330 0.8535 0.5198 0.7977 0.5619 LDT[ 109 ]31.53 28.39 26.70 30.52 26.71 28.20 24.90 0.8977 0.8052 0.7468 0.8399 0.6694 0.7922 0.6358 FCN[ 94 ]33.22 29.49 27.72 32.36 27.67 29.51 25.45 0.9267 0.8599 0.8142 0.8853 0.7340 0.8339 0.6771 IRCNN[ 93 ]34.33 30.04 28.51 33.57 27.64 30.63 25.65 0.9210 0.8156 0.7762 0.8977 0.6884 0.8645 0.6640 FDN[ 87 ]34.05 29.77 27.94 32.63 27.75 29.93 25.93 0.9335 0.8583 0.8139 0.8887 0.7319 0.8555 0.6943 FNBD[ 88 ]34.81 30.63 27.93 31.22 27.63 30.92 25.49 0.9398 0.8658 0.7759 0.8860 0.7010 0.8799 0.6589 RGDN[ 92 ]33.96 29.71 27.45 31.25 26.93 29.51 25.33 0.9395 0.8662 0.7889 0.8869 0.7161 0.8616 0.6688 VEM[ 99 ]34.31 30.50 28.52 32.73 29.41 − − 0.9382 0.8798 0.8348 0.8952 0.8055 − − DWDN[ 101 ]36.90 32.77 30.77 34.05 − 31.74 − 0.9614 0.9179 0.8857 0.9225 − 0.8938 − CV-CNN[ 97 ]35.44 30.85 28.80 33.10 29.54 − − 0.9467 0.8829 0.8381 0.9022 0.8094 − − SVMAP[ 110 ]− − − 34.51 29.20 31.89 27.25 − − − 0.9273 0.7940 0.8973 0.7550
Tools
Get Citation
Copy Citation Text
Hang YANG. Survey of non-blind image restoration[J]. Chinese Optics, 2022, 15(5): 954
Paper Information
Category: Review
Received: May. 16, 2022
Accepted: --
Published Online: Sep. 29, 2022
The Author Email:
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20