基于双通道残差网络的水下图像去噪研究

doi:10.19678/j.issn.1000-3428.0064662

摘要/Abstract

摘要： 水下图像去噪作为探索海底世界有效的辅助手段之一，备受研究人员的关注。传统的滤波方法存在去噪过程中容易损坏图像的细节，带有明显的噪声残留，且根据不同的噪声类型设计相应滤波器的问题，在深度卷积神经网络的基础上，提出一种改进的双通道残差卷积神经网络模型，用于去除水下图像的噪声。该模型由局部残差网（LRN）、全局稀疏网（GSN）和特征处理块（FPB）构成。通过双通道LRN和GSN并行提取水下图像的多层次局部噪声特征和全局噪声特征，利用FPB中的通道连接融合LRN和GSN提取的噪声特征，并使用其卷积层增强水下图像噪声信息。在此基础上，使用均方误差和平均绝对误差双损失函数优化网络参数，利用卷积层重构水下图像。实验结果表明，相比BM3D、IRCNN、DnCNN等方法，该方法的平均峰值信噪比提高0.02~3.52 dB，在有效去除各种水平的随机噪声同时能重构清晰的水下图像。

关键词: 卷积神经网络, 双通道, 残差学习, 图像去噪, 水下图像, 图像处理

Abstract: As an effective underwater exploration tool, underwater image denoising has received increasing attention and research in academia.To address the vulnerability of traditional filtering methods to image degradation in the denoising process, with obvious noise residues, as well as facilitate the design of filters tailored to specific types of noise, an improved dual-channels residual Convolutional Neural Network(CNN) model is proposed herein on the basis of a deep CNN to eliminate noise from underwater images.The proposed model consists of a Local Residual Network(LRN), Global Sparse Network(GSN), and Feature Processing Block(FPB).The multi-level local noise and global noise features of underwater images are extracted via the dual-channels LRN and GSN in parallel.The channel connection in FPB is then used to fuse the extracted noise features from LRN and GSN, and the convolution layer is employed to enhance the noise information.On this basis, two double loss functions-Mean Square Error(MSE) and Mean Absolute Error(MAE)are used to optimize the network parameters, and the convolution layer is used to reconstruct the underwater images.The experimental results show that compared with that of BM3D, IRCNN, DnCNN, and other methods, the average Peak Signal-to-Noise Ratio (PSNR) of the proposed method increases by 0.02~3.52 dB.Thus, the proposed method successfully removes various levels of random noise and reconstructs clear underwater images.

Key words: Convolutional Neural Network(CNN), dual-channels, residual learning, image denoising, underwater images, image processing

中图分类号:

TP391

杨晶晶, 谢海燕, 薛妮妮, 张傲明. 基于双通道残差网络的水下图像去噪研究[J]. 计算机工程, 2023, 49(4): 188-198.

YANG Jingjing, XIE Haiyan, XUE Nini, ZHANG Aoming. Research on Underwater Image Denoising Based on Dual-Channels Residual Network[J]. Computer Engineering, 2023, 49(4): 188-198.

https://www.ecice06.com/CN/Y2023/V49/I4/188

图/表 14

20230417185659

20230417185702

20230417185705

20230417185709

20230417185712

20230417185715

20230417185719

20230417185722

20230417185725

20230417185728

20230417185731

20230417185735

20230417185738

20230417185741

参考文献

[1] 宋清昆, 马丽, 曹建坤, 等.基于小波变换和均值滤波的图像去噪[J].黑龙江大学自然科学学报, 2016, 33(4):555-560. SONG Q K, MA L, CAO J K, et al.Image denoising based on wavelet transform and mean filtering[J].Journal of Natural Science of Heilongjiang University, 2016, 33(4):555-560.(in Chinese)
[2] 张威.空间域中均值滤波与中值滤波去噪的应用研究[J].产业科技创新, 2020(12):65-66. ZHANG W.Application study of mean filtering and median filtering for noise removal in spatial domain[J].Industry Technology Innovation, 2020(12):65-66.(in Chinese)
[3] LEI Z F, SU W B, HU Q.Multimode decomposition and wavelet threshold denoising of mold level based on mutual information entropy[J].Entropy, 2019, 21(2):1-15.
[4] 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.
[5] TIAN C W, XU Y, FEI L K, et al.Enhanced CNN for image denoising[J].CAAI Transactions on Intelligence Technology, 2019, 4(1):17-23.
[6] 周博超, 韩雨男, 桂志国, 等.基于VGG网络与深层字典的低剂量CT图像去噪算法[J].计算机工程, 2022, 48(4):191-196, 205. ZHOU B C, HAN Y N, GUI Z G, et al.Low-dose CT image denoising algorithm based on VGG network and deep dictionary[J].Computer Engineering, 2022, 48(4):191-196, 205.(in Chinese)
[7] JIANG X B, JIN Y, YAO Y.Low-dose CT lung images denoising based on multiscale parallel convolution neural network[J].The Visual Computer, 2021, 37(8):2419-2431.
[8] ZHOU X Y, YANG K D.A denoising representation framework for underwater acoustic signal recognition[J].The Journal of the Acoustical Society of America, 2020, 147(4):377-383.
[9] WANG Y, GUO J, GAO H, et al.UIEC2-Net:CNN-based underwater image enhancement using two color space[J].Signal Processing:Image Communication, 2021, 96:116250.
[10] 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.
[11] HE K M, ZHANG X Y, REN S Q, et al.Deep residual learning for image recognition[C]//Proceedings of Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2016:770-778.
[12] YU F, KOLTUN V.Multi-scale context aggregation by dilated convolutions[EB/OL].[2022-04-03].https://arxiv.org/pdf/1511.07122.pdf.
[13] HE J K, JIANG B, YANG C, et al.Hybrid dilated convolution network using attentive kernels for real-time semantic segmentation[C]//Proceedings of Chinese Conference on Pattern Recognition and Computer Vision.Berlin, Germany:Springer, 2020:129-141.
[14] 姜竣, 翟东海.基于空洞卷积与特征增强的单阶段目标检测算法[J].计算机工程, 2021, 47(7):232-238, 248. JIANG J, ZHAI D H.Single-stage object detection algorithm based on dilated convolution and feature enhancement[J].Computer Engineering, 2021, 47(7):232-238, 248.(in Chinese)
[15] TIAN C W, XU Y, LI Z Y, et al.Attention-guided CNN for image denoising[J].Neural Networks, 2020, 124:117-129.
[16] 周晓杰.空洞卷积神经网络低剂量CT图像去噪[J].现代计算机, 2021(16):160-163. ZHOU X J.Low-dose CT denoising with dilated convolutional neural network[J].Modern Computer, 2021(16):160-163.(in Chinese)
[17] ANWAR S, BARNES N.Real image denoising with feature attention[C]//Proceedings of International Conference on Computer Vision.Washington D.C., USA:IEEE Press, 2020:3155-3164.
[18] JIAO J B, TU W C, HE S F, et al.FormResNet:formatted residual learning for image restoration[C]//Proceedings of Conference on Computer Vision and Pattern Recognition Workshops.Washington D.C., USA:IEEE Press, 2017:1034-1042.
[19] GUO S, YAN Z F, ZHANG K, et al.Toward convolutional blind denoising of real photographs[C]//Proceedings of Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2020:1712-1722.
[20] KINGMA D P, BA J L.Adam:a method for stochastic optimization[EB/OL].[2022-04-03].http://de.arxiv.org/pdf/1412.6980.
[21] ISLAM M J, XIA Y Y, SATTAR J.Fast underwater image enhancement for improved visual perception[J].IEEE Robotics and Automation Letters, 2020, 5(2):3227-3234.
[22] ISLAM M J, LUO P, SATTAR J.Simultaneous enhancement and super-resolution of underwater imagery for improved visual perception[EB/OL].[2022-04-03].https://arxiv.org/abs/2002.01155v1.
[23] LI H Y, LI J J, WANG W.A fusion adversarial underwater image enhancement network with a public test dataset[EB/OL].[2022-04-03].https://arxiv.org/abs/1906.06819v2.
[24] ISLAM M J, EDGE C, XIAO Y Y, et al.Semantic segmentation of underwater imagery:dataset and benchmark[C]//Proceedings of International Conference on Intelligent Robots and Systems.Washington D.C., USA:IEEE Press, 2021:1769-1776.
[25] BEVILACQUA M, ROUMY A, GUILLEMOT C, et al.Low-complexity single-image super-resolution based on nonnegative neighbor embedding[C]//Proceedings of the British Machine Vision Conference.[S.l]:British Machine Vision Association, 2012:1-10.
[26] KODAK E.Kodak lossless true color image suite (PhotoCD PCD0992)[EB/OL].[2022-04-03].http://r0k.us/graphics/kodak.
[27] ZORAN D, WEISS Y.From learning models of natural image patches to whole image restoration[C]//Proceedings of International Conference on Computer Vision.Washington D.C., USA:IEEE Press, 2012:479-486.
[28] HUYNH-THU Q, GHANBARI M.Scope of validity of PSNR in image/video quality assessment[J].Electronics Letters, 2008, 44(13):800.
[29] 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.
[30] JAKUB P, GRZEGORZ M.Based on spectral analysis of voice controllable surveillance system using normalized mean square error[J].Advanced Materials Research, 2011, 340:156-160.
[31] FUKUDA T, KOMATSU S.Image evaluation based on the mean structural similarity for wavefront coding[C]//Proceedings of the 22nd Microoptics Conference.Washington D.C., USA:IEEE Press, 2018:156-157.
[32] 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.
[33] ZHANG K, ZUO W, GU S, et al.Learning deep CNN denoiserprior for image restoration[C]//Proceedings of Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2017:3929-3938.

选择文件类型/文献管理软件名称

选择包含的内容