基于深度学习的去噪技术,通过考虑视觉伪影和整体平滑噪声,提高了图像的质量.然而,它们很少涉及边缘细节的恢复.为此,本文提出了一种基于双域信息的深度残差网络去噪模型,利用小波域信息与空间域信息的融合来扩展网络学习信息,通过在激活单元内引入多尺度学习和空洞卷积,以此提取图像特征,并减少了网络参数.为了进一步改善去噪结果,结合小波域损失和空间域损失构造联合损失函数,使得网络获取更多的边缘与细节.实验结果表明,本文提出的方法不仅可以有效去除图像噪声,而且可以更好地恢复图像纹理细节,在主观和客观评价中均获得了更好的结果.
The denoising technology based on deep learning improves the image quality by considering visual artifacts and overall smoothing noise. However, they rarely involve the restoration of edge details. In this paper, a deep denoising network model based on dual-domain mapping and residual learning is proposed. The network learning information is extended by the fusion of wavelet domain information and spatial domain information. By introducing multi-scale learning and dilated convolution into the activation unit, the image features are extracted and the network parameters are reduced. In order to further improve the denoising results, the joint loss function of wavelet domain loss and spatial domain loss are combined to guide the network to obtain more edges and details. Experimental results show that the proposed method can not only effectively remove image noise, but also better restore image texture details, and obtain better results in both subjective and objective evaluation.
[1] FAN L W, ZHANG F, FAN H, et al. Brief review of image denoising techniques[J]. Vis Comput Ind Biomed Art, 2019, 2(1): 7. DOI: 10.1186/s42492-019-0016-7.
[2] ANCHAL A, BUDHIRAJA S, GOYAL B, et al. An efficient image denoising scheme for higher noise levels using spatial domain filters[J]. Biomed Pharmacol J, 2018, 11(2): 625-634. DOI: 10.13005/bpj/1415.
[3] YUAN Q Q, ZHANG Q, LI J, et al. Hyperspectral image denoising employing a spatial–spectral deep residual convolutional neural network[J]. IEEE Trans Geosci Remote Sens, 2019, 57(2): 1205-1218. DOI: 10.1109/TGRS.2018.2865197.
[4] MITICHE L, HOUDA ADAMOU-MITICHE A B, NAIMI H. Medical image denoising using dual tree complex thresholding wavelet transform[C] //2013 IEEE Jordan Conference on Applied Electrical Engineering and Computing Technologies(AEECT), Amman, Jordan, 2014: 1-5. DOI: 10.1109/AEECT.2013.6716477.
[5] MAGGIONI M, KATKOVNIK V, EGIAZARIAN K, et al. Nonlocal transform-domain filter for volumetric data denoising and reconstruction[J]. IEEE Trans Image Process, 2013, 22(1): 119-133. DOI: 10.1109/TIP.2012.2210725.
[6] BUADES A, COLL B, MOREL J M. Non-local means denoising[J]. Image Process Line, 2011, 1: 208-212. DOI: 10.5201/ipol.2011.bcm_nlm.
[7] CHEN R, LIU X M, ZHAI D M, et al. Depth image denoising via collaborative graph Fourier transform[M] //Communications in Computer and Information Science, Singapore: Springer Singapore, 2018: 128-137. DOI: 10.1007/978-981-10-8108-8_12.
[8] MIRI A, SHARIFIAN S, RASHIDI S, et al. Medical image denoising based on 2D discrete cosine transform via ant colony optimization[J]. Optik, 2018, 156: 938-948. DOI: 10.1016/j.ijleo.2017.12.074.
[9] CHEN H, XU W L, BRODERICK N, et al. An adaptive denoising method for Raman spectroscopy based on lifting wavelet transform[J]. J Raman Spectrosc, 2018, 49(9): 1529-1539. DOI: 10.1002/jrs.5399.
[10] GU S H, ZHANG L, ZUO W M, et al. Weighted nuclear norm minimization with application to image denoising[C] //2014 IEEE Conference on Computer Vision and Pattern Recognition. Columbus, OH, USA, 2014: 2862-2869. DOI: 10.1109/CVPR.2014.366.
[11] DABOV K, FOI A, KATKOVNIK V, et al. Image denoising by sparse 3-D transform-domain collaborative filtering[J]. IEEE Trans Image Process, 2007, 16(8): 2080-2095. DOI: 10.1109/TIP.2007.901238.
[12] KNAUS C, ZWICKER M. Progressive image denoising[J]. IEEE Trans Image Process, 2014, 23(7): 3114-3125. DOI: 10.1109/TIP.2014.2326771.
[13] ELHOSENY M, SHANKAR K. Optimal bilateral filter and Convolutional Neural Network based denoising method of medical image measurements[J]. Measurement, 2019, 143: 125-135. DOI: 10.1016/j.measurement.2019.04.072.
[14] WANG X B, HUANG G M, ZHOU Z W, et al. Radar emitter recognition based on the energy cumulant of short time Fourier transform and reinforced deep belief network[J]. Sensors(Basel), 2018, 18(9): 3103. DOI: 10.3390/s18093103.
[15] BAO L, YANG Z L, WANG S Q, et al. Real image denoising based on multi-scale residual dense block and cascaded U-net with block-connection[C] //2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops(CVPRW). Seattle, WA, USA, 2020: 1823-1831. DOI: 10.1109/CVPRW50498.2020.00232.
[16] FALK T, MAI D, BENSCH R, et al. U-Net: deep learning for cell counting, detection, and morphometry[J]. Nat Methods, 2019, 16(1): 67-70. DOI: 10.1038/s41592-018-0261-2.
[17] DIVAKAR N, BABU R V. Image denoising via CNNs: An Adversarial Approach[C]. 2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops(CVPRW). Honolulu, HI, USA, 2017: 1076-1083. DOI: 10.1109/CVPRW.2017.145.
[18] ZHONG Y, LIU L Z, ZHAO D, et al. A generative adversarial network for image denoising[J]. Multimed Tools Appl, 2020, 79(23): 16517-16529. DOI: 10.1007/s11042-019-7556-x.
[19] SINGH G, MITTAL A, AGGARWAL N. ResDNN: deep residual learning for natural image denoising[J]. IET Image Process, 2020, 14(11): 2425-2434. DOI: 10.1049/iet-ipr.2019.0623.
[20] ZHANG K, ZUO W M, CHEN Y J, et al. Beyond a Gaussian denoiser: residual learning of deep CNN for image denoising[J]. IEEE Trans Image Process, 2017, 26(7): 3142-3155. DOI: 10.1109/TIP.2017.2662206.
[21] LAN R S, ZOU H Z, PANG C, et al. Image denoising via deep residual convolutional neural networks[J]. SIViP, 2021, 15(1): 1-8. DOI: 10.1007/s11760-019-01537-x.
[22] JHA N K, SAINI R, NAG S, et al. E2GC: energy-efficient group convolution in deep neural networks[C] //2020 33rd International Conference on VLSI Design and 2020 19th International Conference on Embedded Systems(VLSID), Bangalore, India, 2020: 155-160. DOI: 10.1109/VLSID49098.2020.00044. (
