Infrared and Laser Engineering, Volume. 53, Issue 7, 20240158(2024)
Image defogging algorithm with transmittance prior and luminance awareness
Figures & Tables(16)
Fig. 1. Image defogging algorithm with transmittance prior and luminance awareness (DTPLA)
Fig. 7. Comparison of this paper's method with existing algorithms on synthesized images
Fig. 8. Comparison between this paper's method and existing algorithms on SOTS (outdoor) dataset
Fig. 9. Comparison of the proposed method with existing algorithms on HSTS dataset
Table 1. Bright area transmittance and non-bright area transmittance
View tableView in ArticleTable 1. Bright area transmittance and non-bright area transmittance
Image size Transmittance (bright spot) Transmittance (non-bright areas) 620 pixel×480 pixel 0.151 0.752
Table 2. Comprehensive assessment results under different weight combinations
View tableView in ArticleTable 2. Comprehensive assessment results under different weight combinations
Performance parameter Value SSIM 85.8% $ {f_1} = 0.7 $ $ {f_1} = 0.6 $ $ {f_1} = 0.5 $ $ {f_1} = 0.4 $ $ {f_1} = 0.3 $ PSNR/dB 39.6 $ {f_2} = 0.3 $ $ {f_2} = 0.4 $ $ {f_2} = 0.5 $ $ {f_2} = 0.6 $ $ {f_2} = 0.7 $ Aggregate score - 71.94 67.32 62.7 58.08 53.46
Table 3. Performance evaluation of this paper's method and existing algorithms on synthetic datasets
View tableView in ArticleTable 3. Performance evaluation of this paper's method and existing algorithms on synthetic datasets
Evaluation criteria Retinex RTDA HEDA Ref.[7] Ref.[12] Ref.[21] Ref.[22] Ref.[23] Proposed PSNR/dB 30.4497 31.7961 32.1257 31.7206 31.9967 34.3786 34.3327 32.4368 39.5799 SSIM 73.83% 65.43% 63.7% 78.14% 70.89% 66.11% 82.17% 82.28% 82.54%
Table 4. Performance evaluation results of each algorithm with the addition of two modules
View tableView in ArticleTable 4. Performance evaluation results of each algorithm with the addition of two modules
Evaluation criteria Retinex RTDA HEDA Ref.[7] Ref.[12] Ref.[21] Ref.[22] Ref.[23] Proposed PSNR/dB 30.4497 31.7961 32.1257 31.7206 31.9967 34.3786 34.3327 32.4368 34.6754 SSIM 73.83% 65.43% 63.7% 78.14% 70.89% 66.11% 82.17% 82.28% 82.37% PSNR（+L）/dB 30.2874 30.5988 32.0852 29.7912 30.7672 32.18 33.46 30.93 33.2456 SSIM（+L） 74.95% 68.89% 66.52% 76.91% 75.48% 69.93% 84.84% 84.76% 86.95% PSNR（+L+G）/dB 35.7602 35.952 36.66 33.83 35.76 36.94 37.87 36.88 39.5799 SSIM（+L+G） 73.25% 65.62% 62.81% 75.97% 69.92% 65.98% 81.79% 81.56% 82.54%
Table 5. Performance evaluation of this paper's method and existing algorithms on SOTS dataset (outdoor)
View tableView in ArticleTable 5. Performance evaluation of this paper's method and existing algorithms on SOTS dataset (outdoor)
Evaluation criteria Retinex RTDA HEDA Ref.[7] Ref.[12] Ref.[21] Ref.[22] Ref.[23] Proposed PSNR/dB 25.0675 27.6538 28.0515 24.2232 27.6977 30.6777 29.4938 29.0985 39.6018 SSIM 70.64% 67.865% 68.87% 62.07% 75.49% 69.57% 83.80% 82.02% 82.84%
Table 6. Performance evaluation of this paper's method and existing algorithms on HSTS dataset
View tableView in ArticleTable 6. Performance evaluation of this paper's method and existing algorithms on HSTS dataset
Evaluation criteria Retinex RTDA HEDA Ref.[7] Ref.[12] Ref.[21] Ref.[22] Ref.[23] Proposed PSNR/dB 24.8968 28.9002 28.1276 24.2767 28.9183 32.2211 30.4813 33.5628 40.7701 SSIM 70.12% 70.49% 68.62% 61.57% 75.66% 63.65% 89.72% 88.65% 88.78%
Table 7. Average running time of different defogging algorithms
View tableView in ArticleTable 7. Average running time of different defogging algorithms
Algorithms Time/s Ref.[7] 0.862 Ref.[12] 0.583 Ref.[22] 2.768 Ref.[23] 0.765 Proposed 0.652
Tools
Get Citation
Copy Citation Text
Dongyang SHI, Sheng HUANG, Huanlin LIU, Junlin ZHANG. Image defogging algorithm with transmittance prior and luminance awareness[J]. Infrared and Laser Engineering, 2024, 53(7): 20240158
Paper Information
Category:
Received: Apr. 10, 2024
Accepted: --
Published Online: Aug. 9, 2024
The Author Email:
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20