| 190 | 3 | 239 |
| 下载次数 | 被引频次 | 阅读次数 |
为了增强图像的去噪效果和同时最大限度的保留图像的细节信息，提出了一种有效的图像去噪算法，该算法引入可变形卷积，并与扩张卷积、普通卷积组成“混合稀疏块”来完成主要的去噪任务。采用双重注意力机制，让网络能够关注到更多的重要信息，从而能挖掘到更深层次下的噪声信息；将网络设计成残差结构并在网络深处使用并行结构，以缓解梯度消失、训练缓慢等问题。实验结果证明，该算法与其他去噪算法相比，去噪表现优异且能更好的保留图像细节信息，拥有更高的数据评价指标，体现了更好的鲁棒性。
Abstract：In order to enhance the denoising effect of the image and preserve the details of the image to the greatest extent, an effective image denoising algorithm is proposed. The algorithm introduces deformable convolution and combines with dilated convolution and ordinary convolution to form a "hybrid sparse" block" to complete the main denoising task. The dual attention mechanism is adopted to allow the network to pay attention to more important information, so that it can dig out the noise information at a deeper level, the network is designed as a residual structure and a parallel structure is used in the depth of the network to alleviate gradient disappearance and training slowness, etc. The experimental results show that compared with other denoising algorithms, the algorithm has excellent denoising performance and can better retain image details, has higher data evaluation indicators, and reflects better robustness.
[1] Tian C W,Zhang Q,Sun G L,et al.FFT Consolidated Sparse And Collaborative Representation for Image Classify-Cation[J].Arabian Journal for Science and Engineering,2018,43(2):741-758.
[2] D Abov K ,Foi A ,Katkovnik V,et al.Image Denoising By Sparse 3-D Transform-Domain Collaborative Filtering[J].IEEE Transactions on Image Process-ing,2007,16(8):2080-2095.
[3] Zhang K,Zuo W M,Chen Y J,et al.Beyond a Gaussian Denoiser:Residual Learning of Deep Cnn for Image Deno-Ising[J].IEEE Transactions on Image Processing,2016,26(7):3142-3155.
[4] Tian C W,Xu Y,Zuo W M.Image Denoising Using Deep cnn With batch renormalization[J].Neural Networks,2-020,8(22):461-473.
[5] Tian C W,Xu Y L,Zuo Y,et al.Attention Guided cnn For Image Denoising[J].Neural Networks,2020,12(24):117-129.
[6] Zhang K,Zuo W M,Zhang L.FFDNet:toward a Fast and Flexible Solution For cnn Based Image Denoising[J].IE-EE Transactions on Image Processing,2017,27(9):4608-4622.
[7] Woo S,Park J,Lee J Y,Et Al.CBAM:Convolution-al Block Attention Module[C]// Computer Vision -ECCV,2018,3-19.
[8] 陈钧荣，林涵阳，陈羽中.基于U-Net融合的保留纹理的图像去噪方法[J].小型微型计算机系统，2021,42(4):791-797.
[9] Dai J F,Qi H Z,Xiong Y W,et al.Deformable Convolutional Networks[C]//2017.IEEE.International.Conferenc- e on Computer Vision (ICCV),2017,764-773.
[10] 盖杉，鲍中运.基于深度学习的高噪声图像去噪算法[J].自动化学报，2020,46(12):2672-2680.
[11] 雷景生，闫晨阳，杨忠光.基于Inception 模块的卷积自编码器图像去噪[J].计算机应用与软件，2021,38(2):2 21-226,322.
[12] Mao S R,Jin R H,Yang C,et al.Denoising of 3d Magnetic Resonance Images using a Residual Encoder-decoer Wasserstein Generative Adversarial Network[J].Medical Image Analysis,2019,55(7):165-180.
[13] Helou M E,Susstrunk S.Blind universal Bayesian Image Denoising with Gaussian Noise Level Learning[J].IEEE Transactions on Image Processing,2020,29(2):4885-4897.
[14] Zoran D,Weiss Y.From Learning Models of Natural Image Patches to Whole Image Restoration[C]/ /2011 Intern-atinal Conference on Computer Vision,2011:479-486.
[15] Schmidt U,Roth S.Shrinkage Fields for Effective Image Restoration[C]//2014 IEEE Conference on Computer Vision and Pattern Recognition,2014:2774-2781.
[16] Li H,Liu F.Image Denoising via Sparse and Redundan trepresentations over Learned Dictionaries[J].IEEE Tran- sactions on Image Processing,2006,15(12):3736-3745.
[17] Gu S H,Zhang L,Zuo W M.Weighted Nuclear Norm Minimization with Application to Image Denoising[C]//2014 IEEE Conference on Computer Vision and Pattern Recognition,2014:2862-2869.
基本信息:
中图分类号:TP391.41
引用信息:
[1]许光宇,付海超.基于可变形卷积和注意力机制的图像去噪算法[J].合肥学院学报(综合版),2022,39(05):79-87.
基金信息:
国家自然科学基金项目“两类噪声背景下的非局部图像去噪研究”(61471004); 安徽理工大学博士基金项目“非局部均值图像滤波方法”(ZX942); 研究生创新基金项目：基于深度学习的图像去噪研究(2021CX2106)资助
2022-08-05
2022
2022-09-13
2022-09-27
2022
1
2022-10-28
2022-10-28