Article Details

Abstract

Addressing lack of physical constraints and edge texture distortion in foggy image dehazing, this paper proposes CMFNet-SR, a dehazing network based on CMFNet that integrates physical priors with edge collaboration. This network incorporates an Edge Constraint Attention Block (ECAB) to enhance details, embeds a dual Physical Prior Module (PPM) to optimize feature modulation, and achieves collaborative optimization through a multi-task loss function. Test results on the RESIDE-SOTS and Haze4K datasets show that on the SOTS dataset, the Peak Signal-to-Noise Ratio (PSNR) reaches 26.19 dB and the Structural Similarity Index (SSIM) is 0.944 0, representing improvements of 2.8% and 3.0% over the baseline, respectively. On the Haze4K dataset, the PSNR reaches 19.72 dB and the SSIM is 0.850 5, indicating improvements of 6.4% and 8.7% over the baseline, respectively. The results demonstrate that the proposed method achieves thorough dehazing, excellent detail and color fidelity, and generalizes well with practical value.

Keywords

image dehazing; edge constraint attention; dual Physical Prior Module; multi-task loss function

References

1. [1] Wang Gaofeng, Zhang Sai, Zhang Yanan, etc Image dehazing method based on improved DehazeNet [J]. Computer System Applications, 2021, 30 (5): 208-213
2. [2] Zheng T Y，Xu T Y，Li X D，et al. Improved AOD-Net Dehazing Algorithm for Target Image [C]//2024 5th International Conference on Computer Engineering and Intelligent Control (ICCEIC)，2024：333-337.
3. [3] Song Y D，He Z Q，Qian H，et al. Vision Transformers for Single Image Dehazing [J].IEEE Transactions on Image Processing，2023：1927-1941.
4. [4] Fan C M，Liu T J，Liu K H. Compound Multi-Branch Feature Fusion for Image Deraindrop [C]//2023 IEEE International Conference on Image Processing (ICIP)，2023：3399-3403.
5. [5] Cheng Yihang, Qiao Zhengyu, Huang Yong, etc Brightness adaptive infrared and visible light image fusion based on Retinex theory (invited) [J]. Advances in Laser and Optoelectronics, 2024, 61 (20): 127-140
6. [6] Wang Yu Research on Image Deblurring Method Based on U-net Network [D]. Huainan: Anhui University of Science and Technology, 2024
7. [7] Zamir S W，Arora A，Khan S，et al. CycleISP: Real Image Restoration via Improved Data Synthesis [C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR)，2020：2693-2702.
8. [8] Ning Hao, Huang Jing, Liu Jun, etc Improved algorithm for dehazing dark channel prior images based on atmospheric scattering model [J]. Journal of Jinan University (Natural Science Edition), 2025, 39 (6): 930-934
9. [9] Qin X，Wang Z L，Bai Y C，et al. FFA-Net: Feature Fusion Attention Network for Single Image Dehazing [C]//Proceedings of the AAAI Conference on Artificial Intelligence，2020，34（7）：11908-11915.
10. [10] Roger Research on end-to-end image dehazing algorithm based on multi feature fusion [D]. Xi'an: Northwestern University, 2022
11. [11] Luo P J，Xiao G Q，Gao X B，et al. LKD-Net: Large Kernel Convolution Network for Single Image Dehazing [C]//2023 IEEE International Conference on Multimedia and Expo (ICME)，2023：1601-1606.
12. [12] Dong H，Pan J S，Xiang L，et al. Multi-Scale Boosted Dehazing Network with Dense Feature Fusion [C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)，2020：2154-2164.
13. [13] Wu R Q，Duan Z P，Guo C L，et al. RIDCP: Revitalizing Real Image Dehazing via High-Quality Codebook Priors [C]//2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR)，2023：22282-22291.

Funding

Yibin University Youth Cultivation Project "Research on Image Enhancement Algorithm Based on Membrane Computing" (2022PY09)

About the authors

Xue Yilan (August 1995-), female, Han ethnicity, from Changzhi, Shanxi, lecturer, master's student, research interests include image dehazing and enhancement; Lei Lang (December 2002-), male, Han ethnicity, from Luzhou, Sichuan, undergraduate student, research interests include image processing and pattern recognition; Qian Jiang (March 1994-), male, Han ethnicity, from Tongling, Anhui, teaching assistant, head of the Data Management and Information Service Section, master's student, research interests include image processing and pattern recognition / university informatization; Xie Qina (October 1995-), female, Han ethnicity, from Yibin, Sichuan, lecturer, master's student, research interests include image restoration techniques.