The dehazing task was divided into two subtasks, content information extraction and edge feature restoration, and a two-branch feature joint dehazing network based on multidimensional collaborative attention was proposed, aiming at the problems that the single-feature extraction network had difficulty in collaboratively enhancing content and edge features in hazy image restoration. In the first branch of the network, densely connected convolutional blocks were constructed to extract multi-level content information from hazy images. In the second branch, cascaded multi-scale residual convolutional blocks were used to restore the texture details. Then the image reconstruction module was used to perform the multi-scale reconstruction on the two types of features, achieving the exchange and enhancement of different feature information, and improving the dehazing effect. Attention mechanism was introduced into the network and attention interactions were modeled in space and pixels to efficiently learn the main features of hazy images. The experimental results show that the proposed network outperforms most existing algorithms in objective metrics on multiple datasets, with high content restoration and complete texture details in dehazing visual effects.
Yan YANG,Lipeng CHAO. A two-branch feature joint dehazing network based on multidimensional collaborative attention. Journal of ZheJiang University (Engineering Science), 2025, 59(6): 1119-1129.
Fig.1Framework of two-branch feature joint dehazing network based on multidimensional collaborative attention
Fig.2Structure of dilated residual dense block （DRDB）
Fig.3Structure of multidimensional collaborative attention（MCA）
Fig.4Structure of cascaded convolutional block（CCB）
Fig.5Structure of multi-scale feature extraction block (MFEB)
模型
PSNR
SSIM
M/106
base
23.251
0.932
0.0202
base-5
23.384
0.931
0.0335
base-7
23.279.
0.929
0.0469
base-e
26.987
0.944
0.0271
base-r
27.920
0.956
0.1021
base-s
24.587
0.943
0.1022
本研究算法
28.369
0.961
0.1022
Tab.1Results of ablation experiments on different modules of proposed method
Fig.6Comparison of local magnification of dehazing effect of ablation experiments
模型
PSNR
SSIM
Model-R
22.251
0.958
Model-P
24.384
0.910
Model-C
26.745
0.934
本研究算法
28.369
0.961
Tab.2Results of ablation experiments with different edge extraction operators
Fig.7Dehazing effect with different edge feature operators
Fig.8Comparison of dehazing effect of different algorithms on SOTS-outdoor
Fig.9Comparison of dehazing effect of different algorithms on SOTS-indoor
Fig.10Comparison of dehazing effect of different algorithms on Haze-RD
Fig.11Comparison of dehazing effect of different algorithms on real hazy images
Fig.12Comparison of edge details of dehazing effect of proposed model
方法
SOTS-indoor
SOTS-outdoor
Haze-RD
PSNR
SSIM
PSNR
SSIM
PSNR
SSIM
DCP
15.745
0.791
15.241
0.745
11.373
0.741
DehazeNet
19.570
0.861
22.078
0.883
14.360
0.763
AOD-Net
16.183
0.823
18.721
0.854
15.222
0.744
GCA-Net
25.652
0.956
27.307
0.902
12.286
0.602
UHD
16.240
0.783
16.154
0.807
14.043
0.778
SGID
18.360
0.907
21.743
0.668
14.473
0.782
DehazeFormer
22.758
0.899
22.572
0.942
14.242
0.780
DEA-Net
27.294
0.907
19.295
0.806
13.218
0.750
本研究算法
23.429
0.910
28.369
0.961
16.023
0.791
Tab.3Comparison of objective evaluations of different methods
方法
t/s
M/106
DCP
2.132(Matlab)
—
DehazeNet
1.532(Matlab)
0.008
AOD-Net
0.017
0.002
GCA-Net
0.163
0.703
SGID
0.814
13.867
DehazeFormer
0.462
2.514
本研究算法
0.103
0.102
Tab.4Comparison of dehazing time and model complexity of different methods
[1]
吴迪, 朱青松 图像去雾的最新研究进展[J]. 自动化学报, 2015, 41 (2): 221- 239 WU Di, ZHU Qingsong The latest research progress of image dehazing[J]. Acta Automatica Sinica, 2015, 41 (2): 221- 239
[2]
TAREL J P, HAUTIÈRE N. Fast visibility restoration from a single color or gray level image [C]// IEEE 12th International Conference on Computer Vision. Kyoto: IEEE, 2009: 2201–2208.
[3]
MENG G, WANG Y, DUAN J, et al. Efficient image dehazing with boundary constraint and contextual regularization [C]// IEEE International Conference on Computer Vision. Sydney: IEEE, 2013: 617–624.
[4]
覃宏超, 李炎炎, 龙伟, 等 基于引导滤波和暗原色先验理论透射率估值的视频去雾算法[J]. 浙江大学学报: 工学版, 2018, 52 (7): 1302- 1309 QIN Hongchao, LI Yanyan, LONG Wei, et al Real-time video dehazing using guided filtering and transmissivity estimated based on dark channel prior theory[J]. Journal of Zhejiang University: Engineering Science, 2018, 52 (7): 1302- 1309
[5]
NAYAR S K, NARASIMHAN S G. Vision in bad weather [C]// 7th IEEE International Conference on Computer Vision. Kerkyra: IEEE, 1999: 820–827.
[6]
NARASIMHAN S G, NAYAR S K. Removing weather effects from monochrome images [C]// IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Kauai: IEEE, 2001: II.
[7]
LIHE Z, HE J, YUAN Q, et al PhDnet: a novel physic-aware dehazing network for remote sensing images[J]. Information Fusion, 2024, 106: 102277
doi: 10.1016/j.inffus.2024.102277
[8]
HE K, SUN J, TANG X. Single image haze removal using dark channel prior [C]// IEEE Conference on Computer Vision and Pattern Recognition. Miami: IEEE, 2009: 1956–1963.
[9]
ZHU Q, MAI J, SHAO L A fast single image haze removal algorithm using color attenuation prior[J]. IEEE Transactions on Image Processing, 2015, 24 (11): 3522- 3533
doi: 10.1109/TIP.2015.2446191
[10]
CAI B, XU X, JIA K, et al DehazeNet: an end-to-end system for single image haze removal[J]. IEEE Transactions on Image Processing, 2016, 25 (11): 5187- 5198
[11]
LI B, PENG X, WANG Z, et al. AOD-net: all-in-one dehazing network [C]// IEEE International Conference on Computer Vision. Venice: IEEE, 2017: 4780–4788.
[12]
HE K, SUN J, TANG X Guided image filtering[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013, 35 (6): 1397- 1409
doi: 10.1109/TPAMI.2012.213
[13]
张登银, 鞠铭烨, 王雪梅 一种基于暗通道先验的快速图像去雾算法[J]. 电子学报, 2015, 43 (7): 1437- 1443 ZHANG Dengyin, JU Mingye, WANG Xuemei A fast image daze removal algorithm using dark channel prior[J]. Acta Electronica Sinica, 2015, 43 (7): 1437- 1443
doi: 10.3969/j.issn.0372-2112.2015.07.029
[14]
JU M, DING C, GUO Y J, et al IDGCP: image dehazing based on gamma correction prior[J]. IEEE Transactions on Image Processing, 2019, 29: 3104- 3118
[15]
ZHANG J, TAO D FAMED-net: a fast and accurate multi-scale end-to-end dehazing network[J]. IEEE Transactions on Image Processing, 2020, 29: 72- 84
doi: 10.1109/TIP.2019.2922837
[16]
REN W, MA L, ZHANG J, et al. Gated fusion network for single image dehazing [C]// IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE, 2018: 3253–3261.
[17]
LIU X, MA Y, SHI Z, et al. GridDehazeNet: attention-based multi-scale network for image dehazing [C]// IEEE/CVF International Conference on Computer Vision. Seoul: IEEE, 2019: 7313–7322.
[18]
DAS S D, DUTTA S. Fast deep multi-patch hierarchical network for nonhomogeneous image dehazing [C]// IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. Seattle: IEEE, 2020: 1994–2001.
[19]
BALLA P K, KUMAR A, PANDEY R A 4-channelled hazy image input generation and deep learning-based single image dehazing[J]. Journal of Visual Communication and Image Representation, 2024, 100: 104099
doi: 10.1016/j.jvcir.2024.104099
[20]
QIN X, WANG Z, BAI Y, et al FFA-net: feature fusion attention network for single image dehazing[J]. AAAI Conference on Artificial Intelligence, 2020, 34 (7): 11908- 11915
doi: 10.1609/aaai.v34i07.6865
[21]
SHAO Y, LI L, REN W, et al. Domain adaptation for image dehazing [C]// IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle: IEEE, 2020: 2805–2814.
[22]
SONG Y, HE Z, QIAN H, et al Vision transformers for single image dehazing[J]. IEEE Transactions on Image Processing, 2023, 32: 1927- 1941
doi: 10.1109/TIP.2023.3256763
[23]
YU Y, ZHANG Y, CHENG Z, et al MCA: multidimensional collaborative attention in deep convolutional neural networks for image recognition[J]. Engineering Applications of Artificial Intelligence, 2023, 126: 107079
doi: 10.1016/j.engappai.2023.107079
[24]
ZHANG Y, TIAN Y, KONG Y, et al. Residual dense network for image super-resolution [C]// IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE, 2018: 2472–2481.
[25]
YAN Q, GONG D, SHI Q, et al. Attention-guided network for ghost-free high dynamic range imaging [C]// IEEE/CVF Conference on Computer Vision and Pattern Recognition. Long Beach: IEEE, 2019: 1751–1760.
[26]
杨爱萍, 刘瑾, 邢金娜, 等 基于内容特征和风格特征融合的单幅图像去雾网络[J]. 自动化学报, 2023, 49 (4): 769- 777 YANG Aiping, LIU Jin, XING Jinna, et al Content feature and style feature fusion network for single image dehazing[J]. Acta Automatica Sinica, 2023, 49 (4): 769- 777
[27]
WANG Z, BOVIK A C, SHEIKH H R, et al Image quality assessment: from error visibility to structural similarity[J]. IEEE Transactions on Image Processing, 2004, 13 (4): 600- 612
doi: 10.1109/TIP.2003.819861
[28]
HE T, ZHANG Z, ZHANG H, et al. Bag of tricks for image classification with convolutional neural networks [C]// IEEE/CVF Conference on Computer Vision and Pattern Recognition. Long Beach: IEEE, 2019: 558–567.
[29]
KINGMA D P, BA J, HAMMAD M M. Adam: a method for stochastic optimization [EB/OL]. (2014−12−22)[2024−02−03]. https://arxiv.org/abs/1412.6980v9.
[30]
LI B, REN W, FU D, et al Benchmarking single image dehazing and beyond[J]. IEEE Transactions on Image Processing, 2018, 1: 492- 505
[31]
MANNOS J, SAKRISON D The effects of a visual fidelity criterion of the encoding of images[J]. IEEE Transactions on Information Theory, 1974, 20 (4): 525- 536
doi: 10.1109/TIT.1974.1055250
[32]
ZHANG Y, DING L, SHARMA G. HazeRD: an outdoor scene dataset and benchmark for single image dehazing [C]// IEEE International Conference on Image Processing. Beijing: IEEE, 2017: 3205–3209.
[33]
CHEN D, HE M, FAN Q, et al. Gated context aggregation network for image dehazing and deraining [C]// IEEE Winter Conference on Applications of Computer Vision. Waikoloa Village: IEEE, 2019: 1375–1383.
[34]
ZHENG Z, REN W, CAO X, et al. Ultra-high-definition image dehazing via multi-guided bilateral learning [C]// IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville: IEEE, 2021: 16180–16189.
[35]
BAI H, PAN J, XIANG X, et al Self-guided image dehazing using progressive feature fusion[J]. IEEE Transactions on Image Processing, 2022, 31: 1217- 1229
doi: 10.1109/TIP.2022.3140609
