《计算机应用》唯一官方网站 ›› 2024, Vol. 44 ›› Issue (3): 916-921.DOI: 10.11772/j.issn.1001-9081.2023030376

• 多媒体计算与计算机仿真 • 上一篇    下一篇

基于全变分正则项展开的迭代去噪网络

侯瑞峰1,2, 张鹏程1,2(), 张丽媛1,2, 桂志国1,2, 刘祎1,2, 张浩文1,2, 王书斌1,2   

  1. 1.中北大学 信息与通信工程学院,太原 030051
    2.生物医学成像与影像大数据山西省重点实验室(中北大学),太原 030051
  • 收稿日期:2023-04-06 修回日期:2023-07-03 接受日期:2023-07-04 发布日期:2023-07-31 出版日期:2024-03-10
  • 通讯作者: 张鹏程
  • 作者简介:侯瑞峰(1998—),男,河北邢台人,硕士研究生,主要研究方向:医学图像重建、医学图像处理
    张丽媛(1990—),女,河北石家庄人,讲师,博士,主要研究方向:放射治疗方案优化、医学图像处理
    桂志国(1972—),男,天津人,教授,博士,主要研究方向:信号与信息处理、放射治疗方案优化、图像重建
    刘祎(1987—),女,河南商丘人,副教授,博士,主要研究方向:医学图像重建、医学图像分析
    张浩文(1998—),男,江苏常州人,硕士研究生,主要研究方向:医学图像重建、医学图像处理
    王书斌(1997—),男,山西晋城人,硕士研究生,主要研究方向:医学图像处理。
  • 基金资助:
    山西省应用基础研究基金资助项目(201901D211246);山西省回国留学人员科研资助项目(2021-111)

Iterative denoising network based on total variation regular term expansion

Ruifeng HOU1,2, Pengcheng ZHANG1,2(), Liyuan ZHANG1,2, Zhiguo GUI1,2, Yi LIU1,2, Haowen ZHANG1,2, Shubin WANG1,2   

  1. 1.College of Information and Communication Engineering,North University of China,Taiyuan Shanxi 030051,China
    2.Shanxi Provincial Key Laboratory for Biomedical Imaging and Big Data (North University of China),Taiyuan Shanxi 030051,China
  • Received:2023-04-06 Revised:2023-07-03 Accepted:2023-07-04 Online:2023-07-31 Published:2024-03-10
  • Contact: Pengcheng ZHANG
  • About author:HOU Ruifeng, born in 1998, M. S. candidate. His research interests include medical image reconstruction, medical image processing.
    ZHANG Liyuan, born in 1990, Ph. D., lecturer. Her research interests include radiotherapy program optimization, medical image processing.
    GUI Zhiguo, born in 1972, Ph. D., professor. His research interests include signal and information processing, radiotherapy program optimization, image reconstruction.
    LIU Yi, born in 1987, Ph. D., associate professor. Her research interests include medical image reconstruction, medical image analysis.
    ZHANG Haowen, born in 1998, M. S. candidate. His research interests include medical image reconstruction, medical image processing.
    WANG Shubin, born in 1997, M. S. candidate. His research interests include medical image processing.
  • Supported by:
    Shanxi Provincial Applied Basis Research Fund(201901D211246);Scientific Research Funding Project for Returned Overseas Students of Shanxi Province(2021-111)

摘要:

针对神经网络训练存在解释能力差以及不稳定问题,提出一种基于CP (Chambolle-Pock)算法求解的全变分(TV)正则项展开去噪网络(CPTV-Net),用于解决低剂量计算机断层扫描(LDCT)图像去噪问题。首先,向L1正则项模型引入TV约束项,以保留图像的结构信息;其次,采用CP算法对去噪模型进行求解并得出具体迭代步骤,保证算法的收敛性;最后,借助浅层卷积神经网络学习线性操作的原始对偶变量迭代公式,用神经网络计算模型的解,并通过收集网络参数优化合并数据。在模拟和真实LDCT数据集上的实验结果表明,与残差编码器-解码器卷积神经网络(REDCNN)、TED-Net(Transformer Encoder-decoder Dilation Network)等五种先进的去噪方法相比,CPTV-Net具有较优的峰值信噪比(PSNR)、结构相似度(SSIM)和视觉信息保真度(VIF)评估值,能生成去噪效果明显和细节保留最为完整的LDCT图像。

关键词: 计算机断层扫描, 模型驱动, 原始对偶算法, 卷积神经网络, 图像去噪

Abstract:

For the shortcomings of poor interpretation ability and instability in neural network training, a Chambolle- Pock (CP) algorithm optimized denoising network based on Total Variational (TV) regularization, CPTV-Net, was proposed to solve the denoising problem of Low-Dose Computed Tomography (LDCT) images. Firstly, the TV constraint term was introduced into the L1 regularization term model to preserve the structural information of the image. Secondly, the CP algorithm was used to solve the denoising model and obtain specific iterative steps to ensure the convergence of the algorithm. Finally, the shallow CNN (Convolutional Neural Network) was used to learn the iterative formula of the primal dual variables of the linear operation. The neural network was used to calculate the solution of the model, and the network parameters were collected to optimize the combined data. The experimental results on simulated and real LDCT datasets show that compared with five advanced denoising methods such as REDCNN (Residual Encoder-Decoder Convolutional Neural Network) and TED-Net (Transformer Encoder-decoder Dilation Network), CPTV-Net has the best Peak Signal-to-Noise Ratio (PSNR), Structural SIMilarity (SSIM), and Visual Information Fidelity (VIF) evaluation values, and can generate LDCT images with significant denoising effect and the most details preserved.

Key words: Computed Tomography (CT), model-driven, primal dual algorithm, Convolutional Neural Network (CNN), image denoising

中图分类号: