阶梯式图像去噪方法

doi:10.13190/j.jbupt.2021-103

北京邮电大学学报 ›› 2022, Vol. 45 ›› Issue (1): 52-57.doi: 10.13190/j.jbupt.2021-103

阶梯式图像去噪方法

王靖¹, 姜竹青¹, 门爱东¹, 郭晓强², 王智康³

1. 北京邮电大学人工智能学院, 北京 100876;
2. 国家广播电视总局广播电视科学研究院, 北京 100866;
3. 莫纳什大学医学护理与健康科学学院, 墨尔本 3800

收稿日期:2021-05-23 出版日期:2022-02-28 发布日期:2021-12-16
通讯作者: 姜竹青(1987—),男,讲师,硕士生导师,邮箱:jiangzhuqing@bupt.edu.cn E-mail:jiangzhuqing@bupt.edu.cn
作者简介:王靖(1997—),女,博士生
基金资助:
国家自然科学基金项目（62002026）

A Ladder-Type Denoising Method

WANG Jing¹, JIANG Zhuqing¹, MEN Aidong¹, GUO Xiaoqiang², WANG Zhikang³

1. School of Artifical Intelligence, Beijing University of Posts and Telecommunications, Beijing 100876, China;
2. Academy of Broadcasting Science, National Radio and Television Administration, Beijing 100866, China;
3. Faculty of Medicine Nursing and Health Sciences, Monash University, Melbourne 3800, Australia

Received:2021-05-23 Online:2022-02-28 Published:2021-12-16

摘要/Abstract

摘要： 为了提高对原始红绿蓝（RAW）和标准红绿蓝（sRGB）2种格式的真实噪声图像的去噪性能，提出了一种基于卷积神经网络的阶梯式图像去噪方法。第一阶段，利用单个通道内空间结构信息对图像各个通道独立去噪，获得初步去噪结果。第二阶段，利用噪声图像在不同通道的相关性进一步去噪，获得增强的去噪结果。在所提方法中引入误差反馈机制来减少采样带来的信息损失；使用密集残差连接模块使得提取到的噪声图像特征能更有效地复用和传播；利用通道注意力使得网络有选择性地增强信息量大的特征，抑制无用特征。将所提方法与常用的其他去噪方法比较，实验结果表明，在达姆施塔特噪声数据集的RAW/sRGB数据集上，所提方法分别达到了49.55 dB和39.55 dB的峰值信噪比（PSNR）；在跨通道数据集达到了39.52 dB的PSNR，较目前绝大多数方法具有更好的性能。

关键词: 卷积神经网络, 图像去噪, 误差反馈, 密集残差连接

Abstract: A ladder-type denoising method is proposed to improve the denoising performance for raw red green blue(RAW) and standard red green blue(sRGB) real-world images. In the first stage, each channel of the noisy image is denoised separately utilizing intra-channel structure information. In the second stage, the inter-channel correlation information of the noisy image is utilized to further denoise the whole image, and the final boosted denoising result is obtained. Error feedback mechanism is introduced to reduce the information loss caused by sampling. Additionally residual dense connection makes features more effective for reuse and propagation; channel attention selectively enhances features with large amount of information and suppress useless features. The proposed method is compared with other denoising algorithms, and the results show that the proposed method achieves 49.55 dB peak signal to noise ratio(PSNR) in RAW images and 39.55 dB PSNR in sRGB images on Darmstadt noise dataset, and 39.52 dB PSNR on cross-channel dataset, which realizes competitive performance in comparison with other denoising algorithms.

Key words: convolution neural network, image denoising, error feedback, residual dense connection

中图分类号:

TP181

王靖, 姜竹青, 门爱东, 郭晓强, 王智康. 阶梯式图像去噪方法[J]. 北京邮电大学学报, 2022, 45(1): 52-57.

WANG Jing, JIANG Zhuqing, MEN Aidong, GUO Xiaoqiang, WANG Zhikang. A Ladder-Type Denoising Method[J]. Journal of Beijing University of Posts and Telecommunications, 2022, 45(1): 52-57.

参考文献

[1] XIE J C, ZHANG D L, XU W L. Wavelet image denoi-sing based on multiscale stochastic process coefficient model[J]. Journal of Tsinghua University (Science and Technology), 2003, 43(9):1222-1225. 谢杰成, 张大力, 徐文立. 基于小波系数多尺度随机过程模型的去噪方法[J]. 清华大学学报(自然科学版), 2003, 43(9):1222-1225.
[2] DABOV K, FOI A, KATKOVNIK V, et al. Color image denoising via sparse 3D collaborative filtering with grou-ping constraint in luminance chrominance space[C]//IEEE International Conference on Image Processing. Piscataway:IEEE Press, 2007:I-313-I-316.
[3] 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 Proce-ssing, 2017, 26(7):3142-3155.
[4] PLÖTZ T, ROTH S. Neural nearest neighbors networks[C]//Proceedings of the 32nd International Conference on Neural Information Processing Systems. New York:Curran Associates Inc, 2018:1095-1106.
[5] REMEZ T, LITANY O, GIRYES R, et al. Class-aware fully convolutional Gaussian and poisson denoising[J]. IEEE Transactions on Image Processing, 2018, 27(11):5707-5722.
[6] GUO S, YAN Z F, ZHANG K, et al. Toward convolutional blind denoising of real photographs[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2019:1712-1722.
[7] YUE Z S, YONG H W, ZHAO Q, et al. Variational denoi-sing network:toward blind noise modeling and removal[C]//Advances in neural information processing systems. New York:Curran Associates, 2019:1690-1701.
[8] HOU Y K, XU J, LIU M X, et al. NLH:a blind pixel-level non-local method for real-world image denoising[J]. IEEE Transactions on Image Processing, 2020, 29:5121-5135.
[9] KRAVITZ D J, SALEEM K S, BAKER C I, et al. The ventral visual pathway:an expanded neural framework for the processing of object quality[J]. Trends in Cognitive Sciences, 2013, 17(1):26-49.
[10] ABDELHAMED A, LIN S, BROWN M S. A high-quality denoising dataset for smartphone cameras[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2018:1692-1700.
[11] PLÖTZ T, ROTH S. Benchmarking denoising algorithms with real photographs[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. Piscataway:IEEE Press, 2017:1586-1595.
[12] NAM S, HWANG Y, MATSUSHITA Y, et al. A holistic approach to cross-channel image noise modeling and its application to image denoising[C]//Computer Vision & Pattern Recognition. Piscataway:IEEE Press, 2016:1683-1691.
[13] 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, 2016, 39(6):1256-1272.
[14] GU S H, ZHANG L, ZUO W M, et al. Weighted nuclear norm minimization with application to image denoising[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. Piscataway:IEEE Press, 2014:2862-2869.

[1]	余佳润王亚峰. 一种低信噪比下基于深度学习的DoA估计方法[J]. 北京邮电大学学报, 2022, 45(6): 119-125.
[2]	叶康张淑军郭淇李辉崔雪红. 基于CM-Transformer的连续手语识别[J]. 北京邮电大学学报, 2022, 45(5): 49-53,78.
[3]	武兴坤, 罗涛, 刘爱军, 杨明, 张文静. 面向主动脉缩窄辅助诊断的神经网络模型设计[J]. 北京邮电大学学报, 2022, 45(4): 1-6.
[4]	武兴坤罗涛刘爱军杨明张文静李剑峰李文秀. 面向主动脉缩窄辅助诊断的神经网络模型设计[J]. 北京邮电大学学报, 2022, 45(4): 1-6.
[5]	许同乐, 孟良, 孔晓佳, 苏元浩, 孙砚飞. 基于EEMD的ICNN故障诊断方法[J]. 北京邮电大学学报, 2022, 45(2): 110-116.
[6]	李晓辉, 王先文, 樊韬, 刘佳文, 万宏杰. 基于卷积神经网络的高性能5G下行同步算法[J]. 北京邮电大学学报, 2022, 45(2): 117-123.
[7]	刘永利, 朱亚孟, 晁浩. 多策略MRFO算法的卷积神经网络超参数优化[J]. 北京邮电大学学报, 2021, 44(6): 83-88,95.
[8]	普运伟, 郭江, 刘涛涛, 吴海潇. 基于多学习单元卷积神经网络的雷达辐射源信号识别[J]. 北京邮电大学学报, 2021, 44(6): 74-82.
[9]	王艺霏, 莫爽, 吴文睿, 范少华, 肖丁. 基于内外卷积网络的网络入侵检测[J]. 北京邮电大学学报, 2021, 44(5): 94-100.
[10]	刘国桢, 陈鸿龙. 卷积记忆图协同过滤[J]. 北京邮电大学学报, 2021, 44(3): 21-26.
[11]	彭鹏, 李卓, 梁纪峰, 马天祥, 刘开华. 支持泛洪攻击检测的命名数据网PIT[J]. 北京邮电大学学报, 2021, 44(2): 61-67.
[12]	曹毅, 黄子龙, 盛永健, 刘晨, 费鸿博. D-2-DenseNet噪音鲁棒的城市音频分类模型[J]. 北京邮电大学学报, 2021, 44(1): 86-91.
[13]	蒲悦逸, 王文涵, 朱强, 陈朋朋. 基于CNN-ResNet-LSTM模型的城市短时交通流量预测算法[J]. 北京邮电大学学报, 2020, 43(5): 9-14.
[14]	谢晓燕, 赵欢, 蒋林. 基于视频数据特性的动态手势识别[J]. 北京邮电大学学报, 2020, 43(5): 91-97.
[15]	韦娟, 顾兴权, 宁方立. 一种基于EEMD的异常声音识别方法[J]. 北京邮电大学学报, 2020, 43(5): 112-117.

阶梯式图像去噪方法

A Ladder-Type Denoising Method

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

本文评价