基于边缘特征融合网络的图像去噪方法

文凯; 季娟; 薛晓; 何如瑾

doi:10.19304/J.ISSN1000-7180.2022.0514

基于边缘特征融合网络的图像去噪方法

doi: 10.19304/J.ISSN1000-7180.2022.0514

文凯^{1, 2},
季娟^{1, 2, ,},
薛晓^{1, 2},
何如瑾^{1, 2}

1.
重庆邮电大学通信与信息工程学院, 重庆 400065
2.
重庆邮电大学数智技术应用研究中心, 重庆 400065

基金项目: 重庆市研究生科研创新项目（CYS295）

详细信息

作者简介:
文凯：男,（1972-）,博士,正高级工程师,硕士生导师.主要从事大数据、计算机视觉和移动通信等方面的研究

薛晓：男,（1996-）,硕士研究生.主要研究领域为图像处理

何如瑾：女,（1998-）,硕士研究生.主要研究领域为视频异常检测

通讯作者:
女,（1997-）,硕士研究生.主要研究领域为图像处理. E-mail：1052443623@qq.com

中图分类号: TP183
计量
- 文章访问数: 7
- HTML全文浏览量: 4
- PDF下载量: 0
- 被引次数: 0
出版历程
- 收稿日期: 2022-08-25
- 修回日期: 2022-10-22

Image denoising method based on edge feature fusion network

WEN Kai^{1, 2},
JI Juan^{1, 2
, ,},
XUE Xiao^{1, 2},
HE Rujin^{1, 2}

1.
School of Communication and Information Engineering, Chongqing University of Posts and Telecommunications, Chongqing 400065, China
2.
Research Center of Digital Intelligence Technology Application, Chongqing University of Posts and Telecommunications, Chongqing 400065, China

摘要

摘要:
目前大多数图像去噪算法在去除噪声的同时,通常会丢失图像的边缘细节信息. 针对这一问题,提出了一种基于边缘特征融合的图像去噪方法. 首先,通过基于Canny算子的边缘提取网络提取图像的边缘信息,由于Canny算子不需要训练,这在很大程度上缩短了去噪时间；其次,通过基于残差密集连接的初去噪网络来保证训练的稳定性以及避免梯度消失,实现图像的初步去噪；最后,通过基于信道与空间注意力机制的融合网络将提取的边缘信息图像与初步去噪图像充分融合,自适应地给相对重要的边缘信息分配更大的权值,对图像的边缘细节进行增强,以得到具有更多边缘信息的清晰图像. 实验结果表明,在BSD68和Set12数据集上,与常见的DnCNN、BM3D等去噪方法相比,所提出去噪方法的平均PSNR比DnCNN分别高出0.13 dB、0.29 dB,比BM3D分别高出0.76 dB、0.82 dB,在视觉效果上看也保留了更多的图像细节,同时去噪速率也大幅度的提高.
- 特征融合 /
- 图像去噪 /
- 边缘提取 /
- 图像融合
Abstract:
Most of the current image denoising algorit hms usually cause the loss of edge detail information of the image while removing noise. Aiming at this problem, an image denoising method based on edge feature fusion is proposed. First,the edge information of the image is extracted by the edge extraction network based on Canny operator. Because the Canny operator does not need to be trained, the denoising time is shortened to a large extent. Secondly, the initial denoising network based on dense residual connections is used to ensure the stability of the training and avoid the disappearance of the gradient to achieve the initial denoising of the image. Finally, through the fusion network based on channel and spatial attention mechanism, the extracted edge information image is fully fused with the preliminary denoised image, and the relatively important edge information is adaptively allocated with more weight, and the edge details of the image are enhanced, so as to get a clear image with more edge information. Experimental results show that on BSD68 and Set12 datasets, compared with the common denoising methods such as DnCNN and BM3D, the average PSNR of the proposed denoising method is 0.13 dB and 0.29 dB higher than DnCNN, and 0.76 dB and 0.82 dB higher than BM3D, respectively. In terms of visual effects, more image details are retained, and the denoising rate is also greatly improved.
- Feature fusion /
- Image denoising /
- Edge extraction /
- Image fusion

HTML全文

图 1 总体网络结构图

Figure 1. Overall network structure diagram

下载: 全尺寸图片幻灯片

图 2 不同高斯滤波器核的可视化结果

Figure 2. Visualization results of different Gaussian filter kernels

下载: 全尺寸图片幻灯片

图 3 （a）基本残差单元 (b)本文残差单元

Figure 3. (a) Basic residual unit (b) Residual unit

下载: 全尺寸图片幻灯片

图 4 通道与空间注意网络（CSAN）

Figure 4. Channel and Spatial Attention Network (CSAN)

下载: 全尺寸图片幻灯片

图 5 模型训练流程

Figure 5. Model training process

下载: 全尺寸图片幻灯片

图 6 $ \sigma $=15时Set68数据集中test045图像上不同方法的去噪结果

Figure 6. Denoising results of different methods on the test045 image in the Set68 dataset when $ \sigma $=15

下载: 全尺寸图片幻灯片

图 7 $ \sigma $=25时Set12数据集中Peppers图像上不同方法的去噪结果

Figure 7. Denoising results of different methods on Peppers image in Set12 dataset when $ \sigma $= 25

下载: 全尺寸图片幻灯片

图 8 医学图像上不同方法的去噪结果对比（$ \sigma $=25）

Figure 8. Comparison of denoising results of different methods on medical images ($ \sigma $=25)

下载: 全尺寸图片幻灯片

表 1 不同去噪方法在BSD68数据集上平均PSNR/dB

Table 1. Average PSNR/dB of different denoising methods on BSD68 dataset

去噪方法	BM3D	WNNM	TNRD	EPLL	DnCNN	IRCNN	Proposed
$ \sigma $=15	31.07	31.37	31.42	31.21	31.72	31.63	31.86
$ \sigma $=25	28.57	28.83	28.92	28.68	29.23	29.15	29.34
$ \sigma $=50	25.62	25.87	25.97	25.67	26.23	26.19	26.37

下载: 导出CSV

表 2 噪声水平为15时不同去噪方法的PSNR/dB

Table 2. PSNR/dB of different denoising methods when the noise level is 15

图像	C.man	House	Peppers	Starfish	Monarch	Airplane	Parrot	Lena	Barbara	Boat	Man	Couple	Average
BM3D	31.91	34.93	32.69	31.14	31.85	31.07	31.37	34.26	33.10	32.13	31.92	32.10	32.37
WNNM	32.17	35.13	32.99	31.82	32.71	31.39	31.62	34.27	33.60	32.27	32.11	32.17	32.70
EPLL	31.85	34.17	32.64	31.13	32.10	31.19	31.42	33.92	31.38	31.93	32.00	31.93	32.14
DnCNN	32.61	34.97	33.30	32.20	33.09	31.70	31.83	34.62	32.64	32.42	32.46	32.47	32.86
IRCNN	32.55	34.89	33.31	32.02	32.82	31.70	31.84	34.53	32.43	32.34	32.40	32.40	32.77
Proposed	32.92	35.33	33.61	32.32	33.32	31.96	32.04	34.82	32.92	32.68	32.58	32.71	33.10

下载: 导出CSV

表 3 噪声水平为25不同去噪方法的PSNR/dB

Table 3. PSNR/dB of different denoising methods when the noise level is 25

图像	C.man	House	Peppers	Starfish	Monarch	Airplane	Parrot	Lena	Barbara	Boat	Man	Couple	Average
BM3D	29.45	32.85	30.16	28.56	29.25	28.42	28.93	32.07	30.71	29.90	29.61	29.71	29.97
WNNM	29.64	33.22	30.42	29.03	29.84	28.69	29.15	32.24	31.24	30.03	29.76	29.82	30.26
EPLL	29.26	32.17	30.17	28.51	29.41	28.61	28.95	31.73	28.61	29.74	29.66	29.53	30.03
DnCNN	30.18	33.06	30.87	29.41	30.28	29.13	29.43	32.44	30.00	30.21	30.10	30.12	30.43
IRCNN	30.08	33.06	30.88	29.27	30.09	29.12	29.47	32.43	29.92	30.17	30.04	30.08	30.38
Proposed	30.46	33.66	31.40	29.60	30.57	29.35	29.62	32.77	30.41	30.52	30.21	30.42	30.74

下载: 导出CSV

表 4 噪声水平为50时不同去噪方法的PSNR/dB

Table 4. PSNR/dB of different denoising methods when the noise level is 50

图像	C.man	House	Peppers	Starfish	Monarch	Airplane	Parrot	Lena	Barbara	Boat	Man	Couple	Average
BM3D	26.13	29.69	26.68	25.04	25.82	25.10	25.90	29.05	27.22	26.78	26.81	26.46	26.72
WNNM	26.45	30.33	26.95	25.44	26.32	25.42	26.14	29.25	27.79	26.97	26.94	26.64	27.05
EPLL	26.10	29.12	26.80	25.12	25.94	25.31	25.95	28.68	24.83	26.74	26.79	26.30	26.47
DnCNN	27.03	30.00	27.32	25.70	26.78	25.87	26.48	29.39	26.22	27.20	27.24	26.90	27.18
IRCNN	26.88	29.96	27.33	25.57	26.61	25.89	26.55	29.40	26.24	27.17	27.17	26.88	27.14
Proposed	27.39	30.76	27.81	25.86	27.02	26.01	26.65	29.63	26.84	27.54	27.24	27.24	27.51

下载: 导出CSV

表 5 两种测试集中比较方法与基准算法对比

Table 5. Comparison of two test set methods and benchmark algorithms

测试集	Set12	BSD68
net-1	30.59	29.32
net-2	30.54	29.36
Proposed	30.63	29.39

下载: 导出CSV

表 6 不同去噪方法的运行时间

Table 6. The running time for different denoising methods

方法	设备	计算量/G	耗时/s
文献[5]	CPU	-	0.590
文献[9]	GPU	1.40	0.024
文献[12]	GPU	2.78	0.050
Proposed	GPU	1.18	0.015

下载: 导出CSV

参考文献(17)

[1]	李雅芝, 车强. 图像去噪在火焰识别中的应用研究[J]. 消防界(电子版),2022,8(11):67-69. DOI: 10.16859/j.cnki.cn12-9204/tu.2022.11.015. LI Y Z, CHE Q. Research on the application of image denoising in flame recognition[J]. Fire Protection (Electronic Edition),2022,8(11):67-69. DOI: 10.16859/j.cnki.cn12-9204/tu.2022.11.015.
[2]	曾宪华, 李彦澄, 高歌, 等. 基于残差编解码器的通道自适应超声图像去噪方法[J]. 电子与信息学报,2022,44(7):2547-2558. DOI: 10.11999/JEIT210331. ZENG X H, LI Y C, GAO G, et al. Channel adaptive ultrasound image denoising method based on residual encoder-decoder networks[J]. Journal of Electronics & Information Technology,2022,44(7):2547-2558. DOI: 10.11999/JEIT210331.
[3]	刘迪, 贾金露, 赵玉卿, 等. 基于深度学习的图像去噪方法研究综述[J]. 计算机工程与应用,2021,57(7):1-13. DOI: 10.3778/j.issn.1002-8331.2011-0341. LIU D, JIA J L, ZHAO Y Q, et al. Overview of image denoising methods based on deep learning[J]. Computer Engineering and Applications,2021,57(7):1-13. DOI: 10.3778/j.issn.1002-8331.2011-0341.
[4]	DABOV K, FOI A, KATKOVNIK V, et al. Image denoising by sparse 3-D transform-domain collaborative filtering[J]. IEEE Transactions on Image Processing,2007,16(8):2080-2095. DOI: 10.1109/TIP.2007.901238.
[5]	LIN T H, HONG H, WU L H. Improved BM3D for real image denoising[C]//2021 13th International Conference on Wireless Communications and Signal Processing (WCSP). Changsha: IEEE, 2021: 1-5.
[6]	谭艳丽, 赵永强, 李国琴. 基于编码稀疏表示的船舶航行图像去噪方法[J]. 舰船科学技术,2021,43(6):37-39. DOI: 10.3404/J.ISSN.1672-7649.2021.3A.013. TAN Y L, ZHAO Y Q, LI G Q. Research on denoising method of ship navigation image based on coded sparse representation[J]. Ship Science and Technology,2021,43(6):37-39. DOI: 10.3404/J.ISSN.1672-7649.2021.3A.013.
[7]	GU S H, ZHANG L, ZUO W M, et al. Weighted nuclear norm minimization with application to image denoising[C]//Proceedings of the 2014 IEEE Conference on Computer Vision and Pattern Recognition. Columbus: IEEE, 2014: 2862-2869.
[8]	WU J F, LEE X D. An improved WNNM algorithm for image denoising[J]. Journal of Physics:Conference Series, 2019,1237(2):022037. DOI: 10.1088/1742-6596/1237/2/022037.
[9]	ZHANG K, ZUO W M, CHEN Y J, et al. Beyond a Gaussian denoiser: Residual learning of deep CNN for image denoising[J]. IEEE Transactions on Image Processing,2017,26(7):3142-3155. DOI: 10.1109/TIP.2017.2662206.
[10]	TIAN C W, XU Y, LI Z Y, et al. Attention-guided CNN for image denoising[J]. Neural Networks,2020,124:117-129. DOI: 10.1016/j.neunet.2019.12.024.
[11]	TIAN C W, XU Y, ZUO W M, et al. Designing and training of a dual CNN for image denoising[J]. Knowledge-Based Systems,2021,226:106949. DOI: 10.1016/j.knosys.2021.106949.
[12]	TIAN C W, XU Y, ZUO W M. Image denoising using deep CNN with batch renormalization[J]. Neural Networks,2020,121:461-473. DOI: 10.1016/j.neunet.2019.08.022.
[13]	郭恒意, 贾振堂. 结合残差密集块的卷积神经网络图像去噪方法[J]. 计算机工程与设计,2020,41(7):1998-2003. DOI: 10.16208/j.issn1000-7024.2020.07.031. GUO H Y, JIA Z T. Convolutional neural network image denoising method combined with residual dense block[J]. Computer Engineering and Design,2020,41(7):1998-2003. DOI: 10.16208/j.issn1000-7024.2020.07.031.
[14]	ZHANG K, ZUO W M, GU S H, et al. Learning deep CNN denoiser prior for image restoration[C]//Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu: IEEE, 2017: 2808-2817.
[15]	CHEN Y J, POCK T. Trainable nonlinear reaction diffusion: A flexible framework for fast and effective image restoration[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2017,39(6):1256-1272. DOI: 10.1109/TPAMI.2016.2596743.
[16]	王润. 图像边缘检测算子的适用场景研究[J]. 电脑知识与技术,2019,15(13):211-214. DOI: 10.14004/j.cnki.ckt.2019.1530. WANG R. Research on the applicable scenarios of image edge detection operators[J]. Computer Knowledge and Technology,2019,15(13):211-214. DOI: 10.14004/j.cnki.ckt.2019.1530.
[17]	HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C]//Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016: 770-778.