多尺度细节增强与分层抑噪的图像去雾算法

doi:10.13229/j.cnki.jdxbgxb.20240538

摘要/Abstract

摘要：

针对现有多数去雾算法存在复原图像细节模糊及噪声放大的问题，提出了一种基于金字塔结构的多尺度细节增强与分层抑噪的图像去雾算法。首先，设计了一种多尺度细节增强算法，将有雾图像通过伽马矫正生成的多幅不同曝光图像加权融合，得到一幅细节增强后的有雾图像以及对应的细节层与模糊层图像，以增强复原图像的细节。其次，构建了一种非局部加权平均算法优化暗直接衰减先验估计的初始透射率，以减少形态学伪影，并利用小半径加权引导滤波（WGIF）进一步细化，求得最终的透射率。最后，根据本文提出的多尺度分层抑噪去雾算法复原无雾图像并抑制噪声放大。实验结果表明，本文算法能更好地抑制噪声放大，得到的无雾图像细节清晰、色彩自然，天空区域复原质量更高，多项客观评价指标相较于当前主流算法显著提升。

关键词: 图像处理, 图像去雾, 金字塔结构, 细节增强, 分层抑噪

Abstract:

An image dehazing algorithm based on a pyramid structure with multi-scale detail enhancement and hierarchical noise suppression is proposed to address the issues of detail blur and noise amplification in existing algorithms. Firstly， a multi-scale detail enhancement algorithm is designed to weight and fuse multiple different exposure images generated by gamma correction， resulting in a fog image after detail enhancement， along with corresponding detail layer and fuzzy layer images， for enhancing the details of the restored image. Secondly， a non-local weighted average algorithm is constructed to optimize the initial transmittance estimated by prior dark direct attenuation， so as to reduce morphological artifacts， while the final transmittance is obtained using a small radius Weighted Guided Image Filter （WGIF）. Finally， through the proposed multi-scale hierarchical noise suppression and fog removal algorithm， the fog-free image is restored while noise amplification is suppressed. Experimental results demonstrate that the proposed algorithm can better suppress noise amplification， producing fog-free images with clear details， natural colors， and higher-quality sky region restoration. Furthermore， multiple objective evaluation metrics are significantly improved compared to those of current mainstream algorithms.

Key words: image processing, image dehazing, pyramid structure, detail enhancement, layered noise suppression

中图分类号:

TP391.4

杨燕,沈汪良. 多尺度细节增强与分层抑噪的图像去雾算法[J]. 吉林大学学报(工学版), 2025, 55(12): 4010-4023.

Yan YANG,Wang-liang SHEN. Multi⁃scale detail enhancement and layered noise suppression algorithm for image dehazing[J]. Journal of Jilin University(Engineering and Technology Edition), 2025, 55(12): 4010-4023.

图/表 15

图1

图2

图3

图4

图5

图6

图7

图8

图9

图10

图11

图12

表1

不同算法客观指标1对比"

算法	指标	（1）	（2）	（3）	（4）	（5）	（6）	（7）	（8）	（9）	（10）	均值
DCP	e	58.347	7.930	29.344	5.998	1.612	1.340	0.297	0.245	45.506	0.444	15.106
	r	6.902	3.057	5.444	4.050	2.114	1.018	2.399	1.380	1.638	2.671	3.067
	s	10.808	12.271	12.007	12.015	13.334	13.493	15.748	14.606	12.600	16.522	13.340
	IVM	6.406	5.317	5.983	4.878	6.609	8.544	9.629	9.627	9.950	10.013	7.696
	VC	54.333	44.833	53.667	20.500	28.333	55.833	96.500	57.500	50.143	84.839	54.648
HLP	e	60.140	7.103	31.861	3.151	2.956	1.416	0.108	0.151	55.699	0.242	16.283
	r	9.662	5.695	7.662	8.056	3.540	2.424	2.496	2.332	3.115	2.399	4.738
	s	10.012	14.119	13.311	13.593	14.470	14.934	16.355	15.941	14.727	17.242	14.470
	IVM	6.466	5.968	5.994	4.808	6.515	9.830	9.830	9.868	11.470	9.404	8.015
	VC	30.667	24.833	24.333	37.167	37.833	69.000	95.833	85.655	44.143	78.871	52.834
AMEF	e	25.399	3.158	4.897	1.838	0.797	1.101	0.370	0.194	50.018	0.458	8.813
	r	2.367	2.532	2.239	2.218	2.284	2.318	1.934	2.240	2.864	1.992	2.299
	s	10.687	12.906	11.938	11.893	14.517	15.383	16.463	16.318	15.208	17.992	14.322
	IVM	0.512	2.324	1.032	1.823	4.505	7.614	9.607	9.230	6.510	10.251	5.341
	VC	16.000	43.333	29.167	40.333	45.500	64.000	96.167	74.167	35.429	75.887	51.998
MLGP	e	33.047	7.615	22.273	5.275	1.175	1.425	0.037	0.137	57.008	0.362	12.835
	r	4.500	3.117	3.951	3.586	2.018	2.176	2.246	2.446	2.167	1.986	2.819
	s	14.035	14.926	14.552	14.342	15.398	15.140	16.302	15.206	14.151	17.622	15.167
	IVM	3.734	6.110	5.301	5.083	6.199	9.663	10.138	9.872	10.790	10.436	7.733
	VC	42.167	53.167	56.333	31.667	37.667	59.667	95.667	91.667	18.714	76.290	56.300
SGID	e	24.907	3.323	3.757	1.472	0.710	1.015	0.345	0.328	45.485	0.363	8.171
	r	1.980	1.752	1.864	2.072	1.398	1.527	1.631	1.607	1.504	1.915	1.725
	s	11.732	12.665	12.168	11.680	13.790	14.078	15.796	14.828	12.355	17.050	13.614
	IVM	0.528	2.556	0.904	1.767	4.432	7.308	9.336	9.613	7.060	8.871	5.237
	VC	26.667	42.500	28.000	44.333	39.500	41.667	74.667	65.625	46.571	60.081	46.961
本文	e	63.434	17.427	35.174	5.815	1.071	1.315	0.153	0.386	59.300	0.350	18.442
	r	12.195	6.332	9.292	6.651	3.304	3.893	4.387	2.710	8.100	2.325	5.919
	s	11.961	13.309	14.738	12.560	13.723	14.519	16.592	15.933	15.700	17.073	14.611
	IVM	4.673	5.395	5.571	5.246	5.703	9.863	10.637	10.161	11.906	10.308	7.946
	VC	63.667	39.833	41.500	42.500	51.000	84.500	100.000	93.750	89.429	81.048	68.723

表1

表2

图13

参考文献 44

[1]	Thejas K, Chandra S. Autonomous bot using machine learning and computer vision[J].SN Computer Science, 2021, 2(4): 251.
[2]	Hassan N, Ullah S, Bhatti N, et al. A cascaded approach for image defogging based on physical and enhancement models[J]. Signal, Image and Video Processing, 2020, 14(5): 867-875.
[3]	Haller I, Nedevschi S. Design of interpolation functions for subpixel-accuracy stereo-vision systems[J]. IEEE Transactions on Image Processing, 2012, 21(2): 889-898.
[4]	Soni B, Mathur P. An improved image dehazing technique using CLAHE and guided filter[C]//2020 7th International Conference on Signal Processing and Integrated Networks (SPIN), Noida, India, 2020:902-907.
[5]	Sarkar M, Sarkar P R, Mondal U, et al. Empirical wavelet transform‐based fog removal via dark channel prior[J]. IET Image Processing, 2020, 14(6): 1170-1179.
[6]	Sankar M R, Krishna P R, Yamini A,et al. Optimization of single image dehazing based on stationary wavelet transform[C]∥Proceedings of the International Conference on Cognitive Computing and Cyber Physical Systems, Hainan, China, 2023: 414-421.
[7]	Su C, Li Z, Wei Z L, et al. Clarity method of low-illumination and dusty coal mine images based on improved Amef[J]. Informatica, 2023, 47: 101-114.
[8]	张驰, 谭南林, 李响, 等. 基于改进型Retinex算法的雾天图像增强技术[J]. 北京航空航天大学学报,2019, 45(2): 309-316.
	Zhang Chi, Tan Nan-lin, Li Xiang, et al. Foggy image enhancement technology based on improved Retinex algorithm[J]. Journal of Beijing University of Aeronautics and Astronsutics, 2019, 45(2): 309-316.
[9]	Wang J B, Lu K, Xue J, et al. Single image dehazing based on the physical model and MSRCR algorithm[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2018, 28(9): 2190-2199.
[10]	Cai B L, Xu X M, 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]	Frants V, Agaian S, Panetta K. QCNN-H: Single-image dehazing using quaternion neural networks[J]. IEEE Transactions on Cybernetics, 2023, 53(9): 5448-5458.
[12]	李硕士, 刘洪瑞, 甘永东, 等. 基于残差密集块与注意力机制的图像去雾网络[J]. 湖南大学学报: 自然科学版, 2021, 48(6): 112-118.
	Li Shuo-shi, Liu Hong-rui, Gan Yong-dong, et al. Image dehazing network based on residual dense block and attention mechanism[J]. Journal of Hunan University (Natural Sciences), 2021, 48(6): 112-118.
[13]	Lyu Z, Chen Y, Hou Y M. MCPNet: Multi-space color correction and features prior fusion for single-image dehazing in non-homogeneous haze scenarios[J]. Pattern Recognition, 2024, 150: No.110290.
[14]	Yi W C, Dong L Q, Liu M, et al. Priors-assisted dehazing network with attention supervision and detail preservation[J]. Neural Networks, 2024, 173: No.106165.
[15]	Wang S B, Mei X S, Kang P S, et al. DFC-dehaze: an improved cycle-consistent generative adversarial network for unpaired image dehazing[J]. The Visual Computer, 2024, 40(4): 2807-2818.
[16]	肖进胜, 申梦瑶, 雷俊锋, 等. 基于生成对抗网络的雾霾场景图像转换算法[J]. 计算机学报, 2020, 43(1): 165-176.
	Xiao Jin-sheng, Shen Meng-yao, Lei Jun-feng, et al. Haze scene image conversion algorithm based on generative adversarial network[J]. Chinese Journal of Computers, 2020, 43(1): 165-176.
[17]	Wang P Y, Zhu H Q, Huang H, et al. TMS-GAN:a twofold multi-scale generative adversarial network for single image dehazing[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2022, 32(5): 2760-2772.
[18]	Zhang J, Dong Q Q, Song W J. GGADN: guided generative adversarial dehazing network[J]. Soft Computing, 2021, 27(3): 1731-1741.
[19]	Wang Y Z, Yan X F, Wang F, et al. UCL-dehaze:toward real-world image dehazing via unsupervised contrastive learning[J]. IEEE Transactions on Image Processing, 2024, 33: 1361-1374.
[20]	Meng G F, Wang Y, Duan J Y, et al.Efficient image dehazing with boundary constraint and contextual regularization[C]∥2013 IEEE International Conference on Computer Vision, Sydney, Australia, 2013: 617-624.
[21]	王柯俨, 胡妍, 王怀, 等. 结合天空分割和超像素级暗通道的图像去雾算法[J]. 吉林大学学报: 工学版, 2019, 49(4): 1377-1384.
	Wang Ke-yan, Hu Yan, Wang Huai, et al. Sky segmentation and super dark channel at pixel level image to fog algorithm[J]. Journal of Jilin University (Engineering and Technology Edition, 2019, 49(4): 1377-1384.
[22]	Kim S E, Park T H, Eom I K. Fast single image dehazing using saturation based transmission map estimation[J]. IEEE Transactions on Image Processing,2019, 29: 1985-1998.
[23]	Liu Y, Shang J X, Pan L, et al. A unified variational model for single image dehazing[J]. IEEE Access,2019, 7: 15722-15736.
[24]	高原原, 胡海苗. 基于多子块协同单尺度Retinex的浓雾图像增强[J]. 北京航空航天大学学报, 2019, 45(5): 944-951.
	Gao Yuan-yuan, Hu Hai-miao. Foggy image enhancement based on multi-block coordinated single-scale Retinex[J]. Journal of Beijing University of Aeronautics and Astronsutics, 2019, 45(5): 944-951.
[25]	Galdran A. Image dehazing by artificial multiple-exposure image fusion[J]. Signal Processing, 2018, 149: 135-147.
[26]	Qu L H, Liu S L, Wang M N, et al. Rethinking multi-exposure image fusion with extreme and diverse exposure levels: A robust framework based on Fourier transform and contrastive learning[J]. Information Fusion, 2023, 92: 389-403.
[27]	Xu K, Wang Q, Xiao H Q, et al.Multi-exposure image fusion algorithm based on Improved weight function[J]. Frontiers in Neurorobotics, 2022, 16: No.846580.
[28]	Yang G, Li J, Gao X B. A dual domain multi-exposure image fusion network based on spatial-frequency integration[J]. Neurocomputing, 2024, 598: No.128146.
[29]	Zhu Q S, Mai J M, Shao L. A fast single image haze removal algorithm using color attenuation prior[J].IEEE Transactions on Image Processing, 2015, 24(11): 3522- 3533.
[30]	Mertens T, Kautz J, Van R F. Exposure fusion[C]∥15th Pacific Conference on Computer Graphics and Applications (PG'07), Tokyo, Japan, 2007: 382-390.
[31]	Hautière N, Tarel J P, Aubert D, et al. Blina contrast enhancement assessment by gradient ratioing at visible edges[J]. Image Analysis and Stereology, 2008, 27(2): 87.
[32]	Nayar S K, Narasimhan S G.Vision in bad weather [C]∥Proceedings of the Seventh IEEE International Conference on Computer Vision, Corinth, Greece, 1999: 820-827.
[33]	Cantor A. Optics of the atmosphere—scattering by molecules and particles[J]. IEEE Journal of Quantum Electronics, 1978, 14(9): 698-699.
[34]	He K M, Sun J, Tang X O. Single image haze removal using dark channel prior[C]∥2009 IEEE Conference on Computer Vision and Pattern Recognition, Miami, USA, 2009: 1956-1963.
[35]	Li Z G, Shu H Y, Zheng C B. Multi-scale single image dehazing using Laplacian and Gaussian Pyramids[J].IEEE Transactions on Image Processing, 2021, 30: 9270-9279.
[36]	Berman D, Treibitz T, Avidan S. Non-local Image dehazing[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, USA, 2016: 1674-1682.
[37]	Li Z G, Zheng J H, Zhu Z J, et al. Weighted guided image filtering[J]. IEEE Transactions on Image Processing, 2015, 24(1): 120-129.
[38]	Li Z G, Zheng J H. Single image de-hazing using globally guided image filtering[J]. IEEE Transactions on Image Processing, 2018, 27(1): 442-450.
[39]	杨燕, 王志伟. 基于补偿透射率和自适应雾浓度系数的图像复原算法[J]. 通信学报, 2020, 41(1):66-75.
	Yang Yan, Wang Zhi-wei. Image restoration algorithm based on compensated transmittance and adaptive haze concentration coefficient[J]. Journal of Communications, 2020, 41(1): 66-75.
[40]	Bai H R, Pan J S, Xiang X G, et al. Self-guided image dehazing using progressive feature fusion[J].IEEE Transactions on Image Processing, 2022, 31: 1217-1229.
[41]	Ancuti C, Ancuti C O, Timofte R, et al. I-HAZE: A dehazing benchmark with real hazy and haze-free indoor images[C]∥Advanced Concepts for Intelligent Vision Systems:19th International Conference, ACIVS 2018, Poitiers, France, 2018: 620-631.
[42]	Li B, Ren W, Fu D, et al. Benchmarking single-image dehazing and beyond[J]. IEEE Transactions on Image Processing, 2018, 28(1): 492-505.
[43]	Huynh-Thu Q, Ghanbari M. Scope of validity of PSNR in image/video quality assessment[J]. Electronics Letters, 2008, 44(13): 800.
[44]	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.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

算法	指标	（11）	（12）	（13）	（14）	（15）	（16）	（17）	（18）	（19）	（20）	（21）	均值
DCP	PSNR	14.787	14.221	9.568	14.669	13.205	15.141	18.878	15.768	18.785	19.902	17.035	15.633
DCP	SSIM	0.775	0.743	0.613	0.732	0.714	0.781	0.822	0.873	0.831	0.836	0.864	0.780
HLP	PSNR	16.064	15.653	17.632	15.197	13.086	16.842	18.102	16.695	16.767	18.303	14.495	16.258
HLP	SSIM	0.823	0.776	0.842	0.773	0.686	0.870	0.853	0.886	0.781	0.826	0.810	0.811
AMEF	PSNR	15.838	18.412	17.756	18.936	15.266	19.902	23.152	18.122	23.034	17.989	16.733	18.649
AMEF	SSIM	0.885	0.852	0.812	0.822	0.887	0.887	0.891	0.826	0.798	0.810	0.762	0.839
MLGP	PSNR	12.491	12.303	16.067	13.520	13.060	16.617	20.026	18.644	19.258	19.388	18.829	16.382
MLGP	SSIM	0.614	0.612	0.794	0.655	0.645	0.863	0.887	0.903	0.841	0.847	0.887	0.777
SGID	PSNR	13.846	12.459	13.659	18.109	15.726	20.713	18.693	18.704	20.055	23.256	16.167	17.399
SGID	SSIM	0.688	0.666	0.668	0.861	0.857	0.917	0.916	0.919	0.942	0.959	0.922	0.847
本文	PSNR	16.452	15.688	20.410	17.740	15.465	19.571	19.377	22.273	21.979	23.416	22.945	19.574
本文	SSIM	0.805	0.854	0.900	0.847	0.849	0.884	0.892	0.901	0.849	0.874	0.892	0.868