Article(id=1325493828240679703, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1325493569112388051, articleNumber=null, orderNo=null, doi=10.7544/issn1000-1239.202550014, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=research-article, receivedDate=1736438400000, receivedDateStr=2025-01-10, revisedDate=1751904000000, revisedDateStr=2025-07-08, acceptedDate=1752681600000, acceptedDateStr=2025-07-17, onlineDate=1793951955136, onlineDateStr=2026-11-06, pubDate=1769875200000, pubDateStr=2026-02-01, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1793951955136, onlineIssueDateStr=2026-11-06, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1793951955136, creator=13701087609, updateTime=1793951955136, updator=13701087609, issue=Issue{id=1325493569112388051, tenantId=1146029695717560320, journalId=1324759914146529478, year='2026', volume='63', issue='2', pageStart='275', pageEnd='538', issueExtLink='null', onlineDate='null', pubDate='1769875200000', pubDateStr='2026-02-01', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=null, createTime=1793951893355, creator='13701087609', updateTime=1793952511329, updator='13701087609', preIssue=null, nextIssue=null, articleTotal=null, ext={EN=IssueExt(id=1325496161167090054, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1325493569112388051, language=EN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=), CN=IssueExt(id=1325496161167090055, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1325493569112388051, language=CN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=)}, issueFiles=null, downloadFileDto=null}, startPage=460, endPage=476, ext={EN=ArticleExt(id=1325493831940055832, articleId=1325493828240679703, tenantId=1146029695717560320, journalId=1324759914146529478, language=EN, title=Stable Image Reconstruction Method Based on Resilient Total Variation Minimization, columnId=1325493703384641624, journalTitle=Journal of Computer Research and Development, columnName=Graphics and Image, runingTitle=null, highlight=null, articleAbstract=

In the field image reconstruction, traditional total variation (TV) regularization often incurs staircase artifacts and compromises image contrast. In this work, we refine the total variation framework by incorporating a non-convex penalty function alongside a backward diffusion mechanism; on this basis, we propose a resilient total variation (RTV)-based image reconstruction model. First, to mitigate staircase artifacts, the proposed method enhances the robustness of TV regularization by incorporating the transformed ${l_1}$ non-convex penalty functions, which improve the model’s adherence to the restricted isometry property (RIP). Second, to mitigate contrast loss in the reconstructed image and suppress noise, a backward diffusion mechanism is incorporated into the model to ensure smoothness in homogeneous regions. Subsequently, based on the RIP conditions, we derive theoretical guarantees for the stable reconstruction of the RTV image reconstruction model. The developed model is solved using a combination scheme of DC (difference of convex functions) programming and the alternating direction multiplier method, yielding an efficient resilient total variation-based reconstruction method. Extensive experiments on standard, natural, and medical images validate the method’s efficacy. The results demonstrate that the proposed method achieves superior reconstruction quality in comparison to the state-of-the-art methods, particularly in under-sampled and noisy scenarios, highlighting its effectiveness, efficiency, and practical utility.

, authors=Feixiang Zhang, Zhihua Li*, Jianjun Liu, authorsList=Feixiang Zhang, Zhihua Li, Jianjun Liu, authorCompany=null, correspAuthors=Zhihua Li, authorNote=

born in 1998. Master candidate. His main research interests include compressed sensing and image reconstruction

born in 1969. PhD, professor. His main research interests include artificial intelligence and its security, cloud computing and cloud data center theory, big data mining, image reconstruction, edge computing, and information security

born in 1987. PhD, associate professor. His main research interests include artificial intelligence, computer vision, and pattern recognition

, correspAuthorsNote=null, copyrightStatement=Copyright © 2026 Journal of Computer Research and Development. All rights reserved., copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=null, magXml=null, pdfUrl=null, pdf=null, pdfFileSize=null, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=null, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=null, mapNumber=null, fund=null), CN=ArticleExt(id=1325493868858319680, articleId=1325493828240679703, tenantId=1146029695717560320, journalId=1324759914146529478, language=CN, title=基于弹性全变分最小化的稳定图像重建方法, columnId=1325493705406296155, journalTitle=计算机研究与发展, columnName=图形与图像, runingTitle=null, highlight=null, articleAbstract=

在图像重建领域,传统的全变分(total variation,TV)会导致重建图像出现阶梯效应和对比度损失。针对此问题,利用非凸罚函数对TV进行改进并引入后向扩散处理机制,基于此,提出一种基于弹性全变分的图像重建模型。首先,为了克服重建图像的阶梯效应,用变形${l_1}$非凸罚函数对TV进行改进,以此加强模型对约束等距性质(restricted isometry property,RIP)的鲁棒性。其次,为了降低重建图像的对比度损失并抑制噪声,通过在模型中引入后向扩散处理机制,以便保证图像均匀区域的平滑性。然后,基于RIP条件推导了模型稳定重建的理论保证。最后,联合DC(difference of convex functions)规划和交替方向乘子法对模型进行求解,基于此进一步提出一个基于弹性全变分的图像重建方法。为了验证所提出方法的有效性、高效性和实用性,分别使用标准图像、自然图像和医学图像进行重建实验。实验结果表明,与已有方法相比,所提出方法在欠采样场景和噪声环境中的重建质量均有不同程度的改善和提高。

, authors=张飞翔, 李志华*, 刘建军, authorsList=张飞翔, 李志华, 刘建军, authorCompany=null, correspAuthors=李志华, authorNote=

1998年生。硕士研究生。主要研究方向为压缩感知、图像重建

1969年生。博士,教授。主要研究方向为人工智能及其安全、云计算与云数据中心理论、大数据挖掘、图像重建、边缘计算、信息安全

1987年生。博士,副教授。主要研究方向为人工智能、计算机视觉、模式识别

, correspAuthorsNote=
李志华()
, copyrightStatement=版权所有 © 2026 计算机研究与发展编辑部, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=fnOtXVbC2+b3ARso18X8UQ==, magXml=T0y0lc808BrpHdlgjeI/0g==, pdfUrl=null, pdf=VUKivJn8fDnZgr6dm/Dk5g==, pdfFileSize=7811866, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=WY111TcmzF2KDHBzBVMkhA==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=HQojbao2zX6goS1V3cBBZQ==, mapNumber=null, fund=null)}, authors=[Author(id=1325493872280871750, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=wszfx@aliyun.com, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1325493873056817992, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, authorId=1325493872280871750, language=EN, stringName=Feixiang Zhang, firstName=Feixiang, middleName=null, lastName=Zhang, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=School of Artificial Intelligence and Computer Science, Jiangnan University, Wuxi, Jiangsu 214122, bio={"content":"

born in 1998. Master candidate. His main research interests include compressed sensing and image reconstruction

"}, bioImg=null, bioContent=

born in 1998. Master candidate. His main research interests include compressed sensing and image reconstruction

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325493873442693961, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, authorId=1325493872280871750, language=CN, stringName=张飞翔, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=江南大学人工智能与计算机学院 江苏无锡 214122, bio={"img":"iq07Ppk9Xnkcg/fHm6tkRQ==","content":"

1998年生。硕士研究生。主要研究方向为压缩感知、图像重建

"}, bioImg=iq07Ppk9Xnkcg/fHm6tkRQ==, bioContent=

1998年生。硕士研究生。主要研究方向为压缩感知、图像重建

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325493871442010946, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, xref=null, ext=[AuthorCompanyExt(id=1325493871823692611, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, companyId=1325493871442010946, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Artificial Intelligence and Computer Science, Jiangnan University, Wuxi, Jiangsu 214122), AuthorCompanyExt(id=1325493871865635652, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, companyId=1325493871442010946, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=江南大学人工智能与计算机学院 江苏无锡 214122)])]), Author(id=1325493875149775691, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=zhli@jiangnan.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1325493876798137166, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, authorId=1325493875149775691, language=EN, stringName=Zhihua Li, firstName=Zhihua, middleName=null, lastName=Li, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=School of Artificial Intelligence and Computer Science, Jiangnan University, Wuxi, Jiangsu 214122, bio={"content":"

born in 1969. PhD, professor. His main research interests include artificial intelligence and its security, cloud computing and cloud data center theory, big data mining, image reconstruction, edge computing, and information security

"}, bioImg=null, bioContent=

born in 1969. PhD, professor. His main research interests include artificial intelligence and its security, cloud computing and cloud data center theory, big data mining, image reconstruction, edge computing, and information security

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325493877775409999, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, authorId=1325493875149775691, language=CN, stringName=李志华, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=江南大学人工智能与计算机学院 江苏无锡 214122, bio={"img":"vqX5uMPZ3EQPkdNkrNDIfQ==","content":"

1969年生。博士,教授。主要研究方向为人工智能及其安全、云计算与云数据中心理论、大数据挖掘、图像重建、边缘计算、信息安全

"}, bioImg=vqX5uMPZ3EQPkdNkrNDIfQ==, bioContent=

1969年生。博士,教授。主要研究方向为人工智能及其安全、云计算与云数据中心理论、大数据挖掘、图像重建、边缘计算、信息安全

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325493871442010946, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, xref=null, ext=[AuthorCompanyExt(id=1325493871823692611, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, companyId=1325493871442010946, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Artificial Intelligence and Computer Science, Jiangnan University, Wuxi, Jiangsu 214122), AuthorCompanyExt(id=1325493871865635652, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, companyId=1325493871442010946, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=江南大学人工智能与计算机学院 江苏无锡 214122)])]), Author(id=1325493878534579026, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1325493878970786644, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, authorId=1325493878534579026, language=EN, stringName=Jianjun Liu, firstName=Jianjun, middleName=null, lastName=Liu, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=School of Artificial Intelligence and Computer Science, Jiangnan University, Wuxi, Jiangsu 214122, bio={"content":"

born in 1987. PhD, associate professor. His main research interests include artificial intelligence, computer vision, and pattern recognition

"}, bioImg=null, bioContent=

born in 1987. PhD, associate professor. His main research interests include artificial intelligence, computer vision, and pattern recognition

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325493881059550037, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, authorId=1325493878534579026, language=CN, stringName=刘建军, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=江南大学人工智能与计算机学院 江苏无锡 214122, bio={"img":"EsvLQpMB4mtaULFIEY6psg==","content":"

1987年生。博士,副教授。主要研究方向为人工智能、计算机视觉、模式识别

"}, bioImg=EsvLQpMB4mtaULFIEY6psg==, bioContent=

1987年生。博士,副教授。主要研究方向为人工智能、计算机视觉、模式识别

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325493871442010946, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, xref=null, ext=[AuthorCompanyExt(id=1325493871823692611, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, companyId=1325493871442010946, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Artificial Intelligence and Computer Science, Jiangnan University, Wuxi, Jiangsu 214122), AuthorCompanyExt(id=1325493871865635652, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, companyId=1325493871442010946, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=江南大学人工智能与计算机学院 江苏无锡 214122)])])], keywords=[Keyword(id=1325493882334618455, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=EN, orderNo=1, keyword=compressed sensing), Keyword(id=1325493882737271640, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=EN, orderNo=2, keyword=image reconstruction), Keyword(id=1325493883135730521, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=EN, orderNo=3, keyword=total variation), Keyword(id=1325493883504829274, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=EN, orderNo=4, keyword=restricted isometry property), Keyword(id=1325493883592909659, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=EN, orderNo=5, keyword=backward diffusion), Keyword(id=1325493883966202716, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=CN, orderNo=1, keyword=压缩感知), Keyword(id=1325493885702644573, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=CN, orderNo=2, keyword=图像重建), Keyword(id=1325493886147240798, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=CN, orderNo=3, keyword=全变分), Keyword(id=1325493886507950943, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=CN, orderNo=4, keyword=约束等距性质), Keyword(id=1325493887103542113, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=CN, orderNo=5, keyword=后向扩散)], refs=[Reference(id=1325493930577503121, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=1, rfOrder=0, authorNames=null, journalName=null, refType=null, unstructuredReference=Donoho D L. Compressed sensing[J]. IEEE Transactions on Information Theory, 2006, 52(4): 1289−1306, articleTitle=null, refAbstract=null), Reference(id=1325493931324089234, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=2, rfOrder=1, authorNames=null, journalName=null, refType=null, unstructuredReference=陈律君, 肖迪, 余柱阳, 等. 基于秘密共享和压缩感知的通信高效联邦学习[J]. 计算机研究与发展, 2022, 59(11): 2395−2407, articleTitle=null, refAbstract=null), Reference(id=1325493931839988627, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=2, rfOrder=2, authorNames=null, journalName=null, refType=null, unstructuredReference=Chen Lüjun, Xiao Di, Yu Zhuyang, et al. Communication-efficient federated learning based on secret sharing and compressed sensing[J]. Journal of Computer Research and Development, 2022, 59(11): 2395−2407 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1325493932288779156, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=3, rfOrder=3, authorNames=null, journalName=null, refType=null, unstructuredReference=Chambolle A, Pock T. An introduction to continuous optimization for imaging[J]. Acta Numerica, 2016, 25(1): 161−319, articleTitle=null, refAbstract=null), Reference(id=1325493933047948181, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=4, rfOrder=4, authorNames=null, journalName=null, refType=null, unstructuredReference=Rudin L I, Osher S, Fatemi E. Nonlinear total variation based noise removal algorithms[J]. Physica D: Nonlinear Phenomena, 1992, 60(1/2/3/4): 259−268, articleTitle=null, refAbstract=null), Reference(id=1325493934725669782, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=5, rfOrder=5, authorNames=null, journalName=null, refType=null, unstructuredReference=Aharon M, Elad M, Bruckstein A. K-SVD: An algorithm for designing overcomplete dictionaries for sparse representation[J]. IEEE Transactions on Signal Processing, 2006, 54(11): 4311−4322, articleTitle=null, refAbstract=null), Reference(id=1325493935996543895, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=6, rfOrder=6, authorNames=null, journalName=null, refType=null, unstructuredReference=Hoerl A E, Kennard R W. Ridge regression: Biased estimation for nonorthogonal problems[J]. Technometrics, 1970, 12(1): 55−67, articleTitle=null, refAbstract=null), Reference(id=1325493936789267352, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=7, rfOrder=7, authorNames=null, journalName=null, refType=null, unstructuredReference=Candès E, Recht B. Exact matrix completion via convex optimization[J]. Communications of the ACM, 2012, 55(6): 111−119, articleTitle=null, refAbstract=null), Reference(id=1325493937238057881, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=8, rfOrder=8, authorNames=null, journalName=null, refType=null, unstructuredReference=Dong Chao, Loy C C, He Kaiming, et al. Image super-resolution using deep convolutional networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 38(2): 295−307, articleTitle=null, refAbstract=null), Reference(id=1325493938072724378, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=9, rfOrder=9, authorNames=null, journalName=null, refType=null, unstructuredReference=Daubechies I, Defrise M, De Mol C. An iterative thresholding algorithm for linear inverse problems with a sparsity constraint[J]. Communications on Pure and Applied Mathematics, 2004, 57(11): 1413−1457, articleTitle=null, refAbstract=null), Reference(id=1325493939331015579, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=10, rfOrder=10, authorNames=null, journalName=null, refType=null, unstructuredReference=Chambolle A. An algorithm for total variation minimization and applications[J]. Journal of Mathematical Imaging and Vision, 2004, 20(1): 89−97, articleTitle=null, refAbstract=null), Reference(id=1325493940127933340, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=11, rfOrder=11, authorNames=null, journalName=null, refType=null, unstructuredReference=Needell D, Ward R. Stable image reconstruction using total variation minimization[J]. SIAM Journal on Imaging Sciences, 2013, 6(2): 1035−1058, articleTitle=null, refAbstract=null), Reference(id=1325493940635444125, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=12, rfOrder=12, authorNames=null, journalName=null, refType=null, unstructuredReference=Bredies K, Kunisch K, Pock T. Total generalized variation[J]. SIAM Journal on Imaging Sciences, 2010, 3(3): 492−526, articleTitle=null, refAbstract=null), Reference(id=1325493941814043550, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=13, rfOrder=13, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhang Honglu, Tang Liming, Fang Zhuang, et al. Nonconvex and nonsmooth total generalized variation model for image restoration[J]. Signal Processing, 2018, 143(1): 69−85, articleTitle=null, refAbstract=null), Reference(id=1325493942678070175, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=14, rfOrder=14, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhu Wei, Chan T. Image denoising using mean curvature of image surface[J]. SIAM Journal on Imaging Sciences, 2012, 5(1): 1−32, articleTitle=null, refAbstract=null), Reference(id=1325493944339014560, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=15, rfOrder=15, authorNames=null, journalName=null, refType=null, unstructuredReference=Fairag F, Chen Ke, Brito-Loeza C, et al. A two-level method for image denoising and image deblurring models using mean curvature regularization[J]. International Journal of Computer Mathematics, 2022, 99(4): 693−713, articleTitle=null, refAbstract=null), Reference(id=1325493944821359521, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=16, rfOrder=16, authorNames=null, journalName=null, refType=null, unstructuredReference=Deng Liangjian, Glowinski R, Tai Xuecheng. A new operator splitting method for the euler elastica model for image smoothing[J]. SIAM Journal on Imaging Sciences, 2019, 12(2): 1190−1230, articleTitle=null, refAbstract=null), Reference(id=1325493945614082978, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=17, rfOrder=17, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhang Hanming, Wang Linyuan, Duan Yuping, et al. Euler’s elastica strategy for limited-angle computed tomography image reconstruction[J]. IEEE Transactions on Nuclear Science, 2017, 64(8): 2395−2405, articleTitle=null, refAbstract=null), Reference(id=1325493946423583651, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=18, rfOrder=18, authorNames=null, journalName=null, refType=null, unstructuredReference=Lou Yifei, Zeng Tieyong, Osher S, et al. A weighted difference of anisotropic and isotropic total variation model for image processing[J]. SIAM Journal on Imaging Sciences, 2015, 8(3): 1798−1823, articleTitle=null, refAbstract=null), Reference(id=1325493948143248292, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=19, rfOrder=19, authorNames=null, journalName=null, refType=null, unstructuredReference=Huo Limei, Chen Wengu, Ge Huanmin, et al. Stable image reconstruction using transformed total variation minimization[J]. SIAM Journal on Imaging Sciences, 2022, 15(3): 1104−1139, articleTitle=null, refAbstract=null), Reference(id=1325493949460259749, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=20, rfOrder=20, authorNames=null, journalName=null, refType=null, unstructuredReference=An Congpei, Wu Haoning, Yuan Xiaoming. Enhanced total variation minimization for stable image reconstruction[J]. Inverse Problems, 2023, 39(7): 075005, articleTitle=null, refAbstract=null), Reference(id=1325493950680802216, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=21, rfOrder=21, authorNames=null, journalName=null, refType=null, unstructuredReference=Alvarez L, Mazorra L. Signal and image restoration using shock filters and anisotropic diffusion[J]. SIAM Journal on Numerical Analysis, 1994, 31(2): 590−605, articleTitle=null, refAbstract=null), Reference(id=1325493953142858665, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=22, rfOrder=22, authorNames=null, journalName=null, refType=null, unstructuredReference=Osher S, Rudin L I. Feature-oriented image enhancement using shock filters[J]. SIAM Journal on Numerical Analysis, 1990, 27(4): 919−940, articleTitle=null, refAbstract=null), Reference(id=1325493953604232106, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=23, rfOrder=23, authorNames=null, journalName=null, refType=null, unstructuredReference=Welk M, Weickert J, Galić I. Theoretical foundations for spatially discrete 1-D shock filtering[J]. Image and Vision Computing, 2007, 25(4): 455−463, articleTitle=null, refAbstract=null), Reference(id=1325493954061411243, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=24, rfOrder=24, authorNames=null, journalName=null, refType=null, unstructuredReference=Gilboa G, Sochen N, Zeevi Y Y. Forward-and-backward diffusion processes for adaptive image enhancement and denoising[J]. IEEE Transactions on Image Processing, 2002, 11(7): 689−703, articleTitle=null, refAbstract=null), Reference(id=1325493955328091052, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=25, rfOrder=25, authorNames=null, journalName=null, refType=null, unstructuredReference=Welk M, Weickert J, Gilboa G. A discrete theory and efficient algorithms for forward-and-backward diffusion filtering[J]. Journal of Mathematical Imaging and Vision, 2018, 60(9): 1399−1426, articleTitle=null, refAbstract=null), Reference(id=1325493957429437359, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=26, rfOrder=26, authorNames=null, journalName=null, refType=null, unstructuredReference=Welk M, Gilboa G, Weickert J. Theoretical foundations for discrete forward-and-backward diffusion filtering[G]//LNCS 5567: Proc of the 2nd Int Conf on Scale Space and Variational Methods in Computer Vision. Berlin: Springer, 2009: 527−538, articleTitle=null, refAbstract=null), Reference(id=1325493957802730416, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=27, rfOrder=27, authorNames=null, journalName=null, refType=null, unstructuredReference=Tikhonov A N, Arsenin V Y. Solutions of Ill-Posed Problems[M]. Hoboken, NJ: John Wiley & Sons, 1977, articleTitle=null, refAbstract=null), Reference(id=1325493958247326641, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=28, rfOrder=28, authorNames=null, journalName=null, refType=null, unstructuredReference=Mizusawa S, Sei Y, Orihara R, et al. Computed tomography image reconstruction using stacked U-Net[J]. Computerized Medical Imaging and Graphics, 2021, 90(1): 101920, articleTitle=null, refAbstract=null), Reference(id=1325493959899882418, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=29, rfOrder=29, authorNames=null, journalName=null, refType=null, unstructuredReference=Liu Jiaming, Sun Yu, Xu Xiaojian, et al. Image restoration using total variation regularized deep image prior[C]//Proc of the 44th Int Conf on Acoustics, Speech and Signal Processing. Piscataway, NJ: IEEE, 2019: 7715−7719, articleTitle=null, refAbstract=null), Reference(id=1325493961552438195, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=30, rfOrder=30, authorNames=null, journalName=null, refType=null, unstructuredReference=Hussein S A, Tirer T, Giryes R. Image-adaptive GAN based reconstruction[C]//Proc of the 34th Int Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2020: 3121−3129, articleTitle=null, refAbstract=null), Reference(id=1325493961988645812, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=31, rfOrder=31, authorNames=null, journalName=null, refType=null, unstructuredReference=Tao P D, An L T H. Convex analysis approach to DC programming: Theory, algorithms and applications[J]. Acta Mathematica Vietnamica, 1997, 22(1): 289−355, articleTitle=null, refAbstract=null), Reference(id=1325493962416464821, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=32, rfOrder=32, authorNames=null, journalName=null, refType=null, unstructuredReference=Boyd S. Distributed optimization and statistical learning via the alternating direction method of multipliers[J]. Foundations and Trends® in Machine Learning, 2010, 3(1): 1−122, articleTitle=null, refAbstract=null), Reference(id=1325493963297268662, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=33, rfOrder=33, authorNames=null, journalName=null, refType=null, unstructuredReference=Chambolle A, Pock T. Learning consistent discretizations of the total variation[J]. SIAM Journal on Imaging Sciences, 2021, 14(2): 778−813, articleTitle=null, refAbstract=null), Reference(id=1325493964563948471, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=34, rfOrder=34, authorNames=null, journalName=null, refType=null, unstructuredReference=Condat L. Discrete total variation: New definition and minimization[J]. SIAM Journal on Imaging Sciences, 2017, 10(3): 1258−1290, articleTitle=null, refAbstract=null), Reference(id=1325493966157783992, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=35, rfOrder=35, authorNames=null, journalName=null, refType=null, unstructuredReference=Bruckstein A M, Donoho D L, Elad M. From sparse solutions of systems of equations to sparse modeling of signals and images[J]. SIAM Review, 2009, 51(1): 34−81, articleTitle=null, refAbstract=null), Reference(id=1325493966589797305, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=36, rfOrder=36, authorNames=null, journalName=null, refType=null, unstructuredReference=Fan Jianqing, Li Runze. Variable selection via nonconcave penalized likelihood and its oracle properties[J]. Journal of the American Statistical Association, 2001, 96(456): 1348−1360, articleTitle=null, refAbstract=null), Reference(id=1325493967063753658, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=37, rfOrder=37, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhang Shuai, Xin J. Minimization of transformed L_1 penalty: Theory, difference of convex function algorithm, and robust application in compressed sensing[J]. Mathematical Programming, 2018, 169(1): 307−336, articleTitle=null, refAbstract=null), Reference(id=1325493968296879035, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=38, rfOrder=38, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhang Shuai, Xin J. Minimization of transformed l_1 penalty: Closed form representation and iterative thresholding algorithms[J]. Communications in Mathematical Sciences, 2017, 15(2): 511−537, articleTitle=null, refAbstract=null), Reference(id=1325493968695337916, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=39, rfOrder=39, authorNames=null, journalName=null, refType=null, unstructuredReference=Sherry F, Benning M, De los Reyes J C, et al. Learning the sampling pattern for MRI[J]. IEEE Transactions on Medical Imaging, 2020, 39(12): 4310−4321, articleTitle=null, refAbstract=null), Reference(id=1325493969991377853, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=40, rfOrder=40, authorNames=null, journalName=null, refType=null, unstructuredReference=Tao P D, An L T H. A DC optimization algorithm for solving the trust-region subproblem[J]. SIAM Journal on Optimization, 1998, 8(2): 476−505, articleTitle=null, refAbstract=null), Reference(id=1325493971639739326, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=41, rfOrder=41, authorNames=null, journalName=null, refType=null, unstructuredReference=Tibshirani R. Regression shrinkage and selection via the lasso[J]. Journal of the Royal Statistical Society Series B: Statistical Methodology, 1996, 58(1): 267−288, articleTitle=null, refAbstract=null), Reference(id=1325493972084335551, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=42, rfOrder=42, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Zhou, 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, articleTitle=null, refAbstract=null), Reference(id=1325493972927390656, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=43, rfOrder=43, authorNames=null, journalName=null, refType=null, unstructuredReference=Weber A. The USC-SIPI image database[EB/OL]. University of Southern California, Signal and Image Processing Institute, 1977[2025-05-05]. http://sipi.usc.edu/database/, articleTitle=null, refAbstract=null), Reference(id=1325493973346821058, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=44, rfOrder=44, authorNames=null, journalName=null, refType=null, unstructuredReference=Lalys F, Haegelen C, Ferre J C, et al. Construction and assessment of a 3-T MRI brain template[J]. NeuroImage, 2010, 49(1): 345−354, articleTitle=null, refAbstract=null), Reference(id=1325493973753668547, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=45, rfOrder=45, authorNames=null, journalName=null, refType=null, unstructuredReference=Gonzalez R C, Woods R E. Digital Image Processing[M]. Noida: Pearson Education India, 2009, articleTitle=null, refAbstract=null)], funds=null, companyList=[AuthorCompany(id=1325493871442010946, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, xref=null, ext=[AuthorCompanyExt(id=1325493871823692611, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, companyId=1325493871442010946, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Artificial Intelligence and Computer Science, Jiangnan University, Wuxi, Jiangsu 214122), AuthorCompanyExt(id=1325493871865635652, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, companyId=1325493871442010946, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=江南大学人工智能与计算机学院 江苏无锡 214122)])], figs=[ArticleFig(id=1325493888613491554, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=EN, label=Figure 1, caption=Parameter optimization analysis via heatmap and contour plot, figureFileSmall=SNGpfMhaD7Paxxz3cHojbw==, figureFileBig=WY111TcmzF2KDHBzBVMkhA==, tableContent=null), ArticleFig(id=1325493890249270115, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=CN, label=图1, caption=基于热力图与等高线图的参数优化分析, figureFileSmall=SNGpfMhaD7Paxxz3cHojbw==, figureFileBig=WY111TcmzF2KDHBzBVMkhA==, tableContent=null), ArticleFig(id=1325493893214643044, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=EN, label=Figure 2, caption=Image data used in experiment, figureFileSmall=wdjGqPTkFV/K019+6Yds/w==, figureFileBig=JpLVcSrq0rxpQ+D3FAtpDw==, tableContent=null), ArticleFig(id=1325493894863004517, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=CN, label=图2, caption=实验用图像数据, figureFileSmall=wdjGqPTkFV/K019+6Yds/w==, figureFileBig=JpLVcSrq0rxpQ+D3FAtpDw==, tableContent=null), ArticleFig(id=1325493895739614054, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=EN, label=Figure 3, caption=Image reconstruction results of Shepp-Logan using RTV method, figureFileSmall=d/4ySZfXsv5MPd/o+RvE3w==, figureFileBig=yj52ogowCt0EwCG0rJzzAQ==, tableContent=null), ArticleFig(id=1325493897002099559, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=CN, label=图3, caption=Shepp-Logan使用RTV方法的图像重建结果, figureFileSmall=d/4ySZfXsv5MPd/o+RvE3w==, figureFileBig=yj52ogowCt0EwCG0rJzzAQ==, tableContent=null), ArticleFig(id=1325493899095057257, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=EN, label=Figure 4, caption=Image reconstruction results of three natural images using RTV method, figureFileSmall=fJ/t820W3drFvIcDNRcYGg==, figureFileBig=0X0uKLXmU78sN4rLEd9S4A==, tableContent=null), ArticleFig(id=1325493899472544618, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=CN, label=图4, caption=3种自然图像使用RTV方法的图像重建结果, figureFileSmall=fJ/t820W3drFvIcDNRcYGg==, figureFileBig=0X0uKLXmU78sN4rLEd9S4A==, tableContent=null), ArticleFig(id=1325493899984249708, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=EN, label=Figure 5, caption=Relationship between evaluation metrics and sampling rates for images reconstructed by RTV method, figureFileSmall=8Dxj9QNosXNxr4V2Gl4mog==, figureFileBig=8InB0qxbKGEWebQ8m8YfkA==, tableContent=null), ArticleFig(id=1325493900714058605, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=CN, label=图5, caption=RTV方法重建图像的评价指标与采样率的关系, figureFileSmall=8Dxj9QNosXNxr4V2Gl4mog==, figureFileBig=8InB0qxbKGEWebQ8m8YfkA==, tableContent=null), ArticleFig(id=1325493901259318126, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=EN, label=Figure 6, caption=Comparison of Shepp-Logan image reconstructed by different methods, figureFileSmall=nTE2+BN07/T6BF+y7ddCWA==, figureFileBig=BbDj7goGNKxgMPyGMg/E0Q==, tableContent=null), ArticleFig(id=1325493902408557424, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=CN, label=图6, caption=各方法重建Shepp-Logan图像对比, figureFileSmall=nTE2+BN07/T6BF+y7ddCWA==, figureFileBig=BbDj7goGNKxgMPyGMg/E0Q==, tableContent=null), ArticleFig(id=1325493903322915698, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=EN, label=Figure 7, caption=Comparison of Cameraman image reconstructed by different methods, figureFileSmall=HlbNcqRxwPfjvOHSOFsWAA==, figureFileBig=C3jwhG9z00TUrmdVVX4XpQ==, tableContent=null), ArticleFig(id=1325493903708791667, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=CN, label=图7, caption=各方法重建Cameraman图像对比, figureFileSmall=HlbNcqRxwPfjvOHSOFsWAA==, figureFileBig=C3jwhG9z00TUrmdVVX4XpQ==, tableContent=null), ArticleFig(id=1325493904119833460, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=EN, label=Figure 8, caption=Comparison of Clock image reconstructed by different methods, figureFileSmall=vzI1CgW2qauOKe2M0WSqWA==, figureFileBig=tL9zEfLUP3XZqGgcNhf3tQ==, tableContent=null), ArticleFig(id=1325493905805943669, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=CN, label=图8, caption=各方法重建Clock图像对比, figureFileSmall=vzI1CgW2qauOKe2M0WSqWA==, figureFileBig=tL9zEfLUP3XZqGgcNhf3tQ==, tableContent=null), ArticleFig(id=1325493906611250038, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=EN, label=Figure 9, caption=Comparison of Peppers image reconstructed by different methods, figureFileSmall=0jGADnO1dYccqLvtyVi6YQ==, figureFileBig=dFP0YTmTezAW1abBCngFLA==, tableContent=null), ArticleFig(id=1325493907932455799, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=CN, label=图9, caption=各方法重建Peppers图像对比, figureFileSmall=0jGADnO1dYccqLvtyVi6YQ==, figureFileBig=dFP0YTmTezAW1abBCngFLA==, tableContent=null), ArticleFig(id=1325493908746150776, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=EN, label=Figure 10, caption=Evaluation metrics for Shepp-Logan image and natural images reconstructed by different methods, figureFileSmall=d7Aw3pwGh2mh2a2t4NU8oQ==, figureFileBig=vgQY3KxzCZoK6NQ18Ao8sQ==, tableContent=null), ArticleFig(id=1325493909589205881, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=CN, label=图10, caption=不同方法对Shepp-Logan图像和自然图像的重建评价指标, figureFileSmall=d7Aw3pwGh2mh2a2t4NU8oQ==, figureFileBig=vgQY3KxzCZoK6NQ18Ao8sQ==, tableContent=null), ArticleFig(id=1325493913317942138, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=EN, label=Figure 11, caption=Relationship between evaluation metrics for image reconstructed by different methods and noise intensity, figureFileSmall=7YDrKqzoU5j2HJHq4e7YJw==, figureFileBig=OHS63rI8f1C9KP3X6a6ncw==, tableContent=null), ArticleFig(id=1325493913716401019, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=CN, label=图11, caption=各方法重建图像的评价指标与噪声强度的关系, figureFileSmall=7YDrKqzoU5j2HJHq4e7YJw==, figureFileBig=OHS63rI8f1C9KP3X6a6ncw==, tableContent=null), ArticleFig(id=1325493914156802941, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=EN, label=Figure 12, caption=Comparison of Brain image reconstructed by different methods, figureFileSmall=v3GLYhIAd6uZ02JGKO+bfw==, figureFileBig=1WeumPWbb9Tjo8qc8NU3xQ==, tableContent=null), ArticleFig(id=1325493914559456126, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=CN, label=图12, caption=各方法重建Brain图像对比, figureFileSmall=v3GLYhIAd6uZ02JGKO+bfw==, figureFileBig=1WeumPWbb9Tjo8qc8NU3xQ==, tableContent=null), ArticleFig(id=1325493914655925119, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=EN, label=Figure 13, caption=Comparison of Spine image reconstructed by different methods, figureFileSmall=uGDf7yCT24+y6mo2EW9c6g==, figureFileBig=4qwaoqvaPYvSmtsSMS8Nbg==, tableContent=null), ArticleFig(id=1325493916690162560, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=CN, label=图13, caption=各方法重建Spine图像对比, figureFileSmall=uGDf7yCT24+y6mo2EW9c6g==, figureFileBig=4qwaoqvaPYvSmtsSMS8Nbg==, tableContent=null), ArticleFig(id=1325493917084427138, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=EN, label=Figure 14, caption=Comparison of objective evaluation parameters for Brain image reconstructed by different methods, figureFileSmall=4/DX58VI4Kgpw+exReFxsQ==, figureFileBig=cLyzqLGf3+5dYllz/q0LsQ==, tableContent=null), ArticleFig(id=1325493917533217667, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=CN, label=图14, caption=不同方法重建Brain图像的客观评价参数对比, figureFileSmall=4/DX58VI4Kgpw+exReFxsQ==, figureFileBig=cLyzqLGf3+5dYllz/q0LsQ==, tableContent=null), ArticleFig(id=1325493917935870852, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=EN, label=Figure 15, caption=Comparison of objective evaluation parameters for Spine image reconstructed by different methods, figureFileSmall=p20MidBXapiteTst+kP7eA==, figureFileBig=FW82weAIJRvkurMjGWU6jA==, tableContent=null), ArticleFig(id=1325493918766343045, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=CN, label=图15, caption=不同方法重建Spine图像的客观评价参数对比, figureFileSmall=p20MidBXapiteTst+kP7eA==, figureFileBig=FW82weAIJRvkurMjGWU6jA==, tableContent=null), ArticleFig(id=1325493919219327878, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=EN, label=Table 1, caption=Parameter Settings for RTV Method, figureFileSmall=null, figureFileBig=null, tableContent=
参数$\alpha $$\beta $$\sigma $$\mu $
取值10.431086000
), ArticleFig(id=1325493921316479879, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=CN, label=表1, caption=

RTV方法参数设置

, figureFileSmall=null, figureFileBig=null, tableContent=
参数$\alpha $$\beta $$\sigma $$\mu $
取值10.431086000
), ArticleFig(id=1325493922146952072, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=EN, label=Table 2, caption=Image Reconstruction Metrics of Shepp-Logan Using RTV Method, figureFileSmall=null, figureFileBig=null, tableContent=
RESSIMPSNR/dBRMSE
7.52×10−81.0000154.650.0000
), ArticleFig(id=1325493922784486281, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=CN, label=表2, caption=

Shepp-Logan使用RTV方法的图像重建指标

, figureFileSmall=null, figureFileBig=null, tableContent=
RESSIMPSNR/dBRMSE
7.52×10−81.0000154.650.0000
), ArticleFig(id=1325493923782730634, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=EN, label=Table 3, caption=Model Parameter Settings for RTV in Natural Image Reconstruction, figureFileSmall=null, figureFileBig=null, tableContent=
模型参数CameramanClockPeppers
$\alpha $0.10.050.5
$\beta $1.21.51.5
), ArticleFig(id=1325493924260881291, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=CN, label=表3, caption=

自然图像重建中RTV的模型参数设置

, figureFileSmall=null, figureFileBig=null, tableContent=
模型参数CameramanClockPeppers
$\alpha $0.10.050.5
$\beta $1.21.51.5
), ArticleFig(id=1325493926009906060, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=EN, label=Table 4, caption=Image Reconstruction Metrics of Three Natural Images Using RTV Method, figureFileSmall=null, figureFileBig=null, tableContent=
图像RESSIMPSNR/dBRMSE
Cameraman0.09590.786625.94700.0504
Clock0.04650.881229.00160.0355
Peppers0.13050.728624.32700.0608
), ArticleFig(id=1325493926727132045, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=CN, label=表4, caption=

3种自然图像使用RTV方法的图像重建指标

, figureFileSmall=null, figureFileBig=null, tableContent=
图像RESSIMPSNR/dBRMSE
Cameraman0.09590.786625.94700.0504
Clock0.04650.881229.00160.0355
Peppers0.13050.728624.32700.0608
), ArticleFig(id=1325493927582770062, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=EN, label=Table 5, caption=Model Parameter Settings for RTV in Medical Image Reconstruction, figureFileSmall=null, figureFileBig=null, tableContent=
模型参数BrainSpine
$\alpha $0.050.1
$\beta $0.140.01
), ArticleFig(id=1325493928400659343, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, language=CN, label=表5, caption=

医学图像重建中RTV的模型参数设置

, figureFileSmall=null, figureFileBig=null, tableContent=
模型参数BrainSpine
$\alpha $0.050.1
$\beta $0.140.01
)], attaches=null, journal=Journal(id=1324759606947315908, delFlag=0, nameCn=计算机研究与发展, nameEn=Journal of Computer Research and Development, nameHistory1=null, nameHistory2=null, issn=1000-1239, eissn=null, cn=11-1777/TP, coden=null, periodic=0, language=CN, oaType=null, ccby=null, superviseOffice=null, ownerOffice=null, pubOffice=null, editorOffice=null, officeType=null, aims=null, clcCode=null, officeProv=null, officeCity=null, officeAddr=null, officeZip=null, officeEmail=null, officePhone=null, editDirector=null, officeDirector=null, officeDirectorPhone=null, officeStaffNum=null, officeEmpNum=null, coverPicUrl=k2l8zgqwstuxxevFX2qxVA==, journalPrice=null, startedYear=null, abbrevIsoEn=Journal of Computer Research and Development, journalRemark=null, publicationField=null, createdTime=1793776903141, updatedTime=1793850698612, createdBy=18614031015, updatedBy=13701087609, firstLetterCn=J, firstLetterEn=J, subjectCode=Electronic Information, subjectName=null, subjectCodeEn=Electronic Information, subjectNameEn=null, picCn=k2l8zgqwstuxxevFX2qxVA==, picEn=r6k0Ry+RD+c/VXj5QarmDw==, jcr=null, cjcr=null, exts=[JournalExt(id=1325069127710286688, language=CN, name=计算机研究与发展, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1793850698640, updatedTime=1793850698640, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://crad.ict.ac.cn/journalx/authorLogOn.action, submissionEditorUrl=https://crad.ict.ac.cn/journalx/editorLogOn.action, submissionReviewUrl=https://crad.ict.ac.cn/journalx/expertLogOn.action, submissionCeEditorUrl=https://crad.ict.ac.cn/journalx/editorInChiefLogOn.action, submissionAeEditorUrl=, option={"copyright":""}), JournalExt(id=1325069127777395553, language=EN, name=Journal of Computer Research and Development, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1793850698656, updatedTime=1793850698656, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://crad.ict.ac.cn/journalx/authorLogOn.action, submissionEditorUrl=https://crad.ict.ac.cn/journalx/editorLogOn.action, submissionReviewUrl=https://crad.ict.ac.cn/journalx/expertLogOn.action, submissionCeEditorUrl=https://crad.ict.ac.cn/journalx/editorInChiefLogOn.action, submissionAeEditorUrl=, option={"copyright":""})], databaseList=null, tenantJournalId=1324759914146529478, websiteList=[Website(id=1325070300139892771, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1324759914146529478, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN, language=CN, createTime=1793850978169, createBy=18614031015, updateTime=1793851040259, updateBy=18614031015, name=计算机研究与发展-中文, tplId=1146099689490845704, title=计算机研究与发展, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1325070893751349360, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=articleTextType, value=kx, createTime=1793851119697, updateTime=1793851119697, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893730377837, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=banner, value=null, createTime=1793851119692, updateTime=1793851119692, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893772320883, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=grayFlag, value=0, createTime=1793851119702, updateTime=1793851119702, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893721989228, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=logo, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/file/pic?fileId=U27tDsy717YF7XSPA+Vwdw==, createTime=1793851119690, updateTime=1793851119690, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893784903797, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=minRunFlag, value=0, createTime=1793851119705, updateTime=1793851119705, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893742960751, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/file/pic, createTime=1793851119695, updateTime=1793851119695, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893780709492, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=silenceFlag, value=0, createTime=1793851119704, updateTime=1793851119704, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893738766446, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1793851119694, updateTime=1793851119694, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893755543665, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=themeColor, value=null, createTime=1793851119698, updateTime=1793851119698, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893763932274, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=themeStyle, value=null, createTime=1793851119700, updateTime=1793851119700, creator=18614031015, updator=18614031015)]), Website(id=1325070300580294698, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1324759914146529478, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN, language=EN, createTime=1793850978274, createBy=18614031015, updateTime=1793851036638, updateBy=18614031015, name=计算机研究与发展-英文, tplId=1146101810881728533, title=Journal of Computer Research and Development, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1325070925288321148, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=articleTextType, value=kx, createTime=1793851127216, updateTime=1793851127216, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925263155321, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=banner, value=null, createTime=1793851127210, updateTime=1793851127210, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925309292671, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=grayFlag, value=0, createTime=1793851127221, updateTime=1793851127221, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925254766712, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=logo, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/file/pic?fileId=U27tDsy717YF7XSPA+Vwdw==, createTime=1793851127208, updateTime=1793851127208, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925326069889, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=minRunFlag, value=0, createTime=1793851127225, updateTime=1793851127225, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925279932539, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/file/pic, createTime=1793851127214, updateTime=1793851127214, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925317681280, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=silenceFlag, value=0, createTime=1793851127223, updateTime=1793851127223, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925271543930, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_en_623/, createTime=1793851127212, updateTime=1793851127212, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925292515453, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=themeColor, value=null, createTime=1793851127217, updateTime=1793851127217, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925300904062, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=themeStyle, value=null, createTime=1793851127219, updateTime=1793851127219, creator=18614031015, updator=18614031015)])], journalTitle=计算机研究与发展, weixinUrl=null, journalUrl=https://crad.ict.ac.cn/, iacademicId=null, status=1, seqNo=null, journalTitleEn=Journal of Computer Research and Development, journalPhotoCn=k2l8zgqwstuxxevFX2qxVA==, journalPhotoEn=r6k0Ry+RD+c/VXj5QarmDw==, journalFirstLetter=J, journalRecommend=null, journalNew=null, journalCollection=null, jcrJf=null, cjcrJf=null, jcrJfStr=null, cjcrJfStr=null, submissionFirstDecision=null, sciSubjectClassification=null, casSubjectClassification=null, citeScore=null, totalCitationFrequency=null, icpCode=null, psCode=null, advertisingLicenseCode=null, copyrightInformation=null, country=null, option=, provinceCode=null, provinceName=null, collectFlag=false, interPubPlatform=, interPubPlatformUrl=null), detailUrlCn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/10.7544/issn1000-1239.202550014, detailUrlEn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/10.7544/issn1000-1239.202550014, pdfUrlCn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/PDF/10.7544/issn1000-1239.202550014, pdfUrlEn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/PDF/10.7544/issn1000-1239.202550014, aliStartDate=0, aliEndDate=0, collectionFlag=false, citedCount=null, citedUrl=null, previewStatus=0, delFlag=0, hasFullText=1, orderTime=1769875200000, fullTextJson=null, articleText=null, reference=null)
收藏切换
基于弹性全变分最小化的稳定图像重建方法
张飞翔 , 李志华 * , 刘建军
计算机研究与发展 | 图形与图像 2026,63(2): 460-476
收起
收藏切换
计算机研究与发展 |图形与图像 2026 , 63 (2) : 460 -476
基于弹性全变分最小化的稳定图像重建方法
全屏
[Author(id=1325493872280871750, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=wszfx@aliyun.com, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1325493873056817992, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, authorId=1325493872280871750, language=EN, stringName=Feixiang Zhang, firstName=Feixiang, middleName=null, lastName=Zhang, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=School of Artificial Intelligence and Computer Science, Jiangnan University, Wuxi, Jiangsu 214122, bio={"content":"

born in 1998. Master candidate. His main research interests include compressed sensing and image reconstruction

"}, bioImg=null, bioContent=

born in 1998. Master candidate. His main research interests include compressed sensing and image reconstruction

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325493873442693961, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, authorId=1325493872280871750, language=CN, stringName=张飞翔, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=江南大学人工智能与计算机学院 江苏无锡 214122, bio={"img":"iq07Ppk9Xnkcg/fHm6tkRQ==","content":"

1998年生。硕士研究生。主要研究方向为压缩感知、图像重建

"}, bioImg=iq07Ppk9Xnkcg/fHm6tkRQ==, bioContent=

1998年生。硕士研究生。主要研究方向为压缩感知、图像重建

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325493871442010946, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, xref=null, ext=[AuthorCompanyExt(id=1325493871823692611, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, companyId=1325493871442010946, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Artificial Intelligence and Computer Science, Jiangnan University, Wuxi, Jiangsu 214122), AuthorCompanyExt(id=1325493871865635652, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, companyId=1325493871442010946, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=江南大学人工智能与计算机学院 江苏无锡 214122)])]), Author(id=1325493875149775691, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=zhli@jiangnan.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1325493876798137166, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, authorId=1325493875149775691, language=EN, stringName=Zhihua Li, firstName=Zhihua, middleName=null, lastName=Li, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=School of Artificial Intelligence and Computer Science, Jiangnan University, Wuxi, Jiangsu 214122, bio={"content":"

born in 1969. PhD, professor. His main research interests include artificial intelligence and its security, cloud computing and cloud data center theory, big data mining, image reconstruction, edge computing, and information security

"}, bioImg=null, bioContent=

born in 1969. PhD, professor. His main research interests include artificial intelligence and its security, cloud computing and cloud data center theory, big data mining, image reconstruction, edge computing, and information security

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325493877775409999, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, authorId=1325493875149775691, language=CN, stringName=李志华, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=江南大学人工智能与计算机学院 江苏无锡 214122, bio={"img":"vqX5uMPZ3EQPkdNkrNDIfQ==","content":"

1969年生。博士,教授。主要研究方向为人工智能及其安全、云计算与云数据中心理论、大数据挖掘、图像重建、边缘计算、信息安全

"}, bioImg=vqX5uMPZ3EQPkdNkrNDIfQ==, bioContent=

1969年生。博士,教授。主要研究方向为人工智能及其安全、云计算与云数据中心理论、大数据挖掘、图像重建、边缘计算、信息安全

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325493871442010946, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, xref=null, ext=[AuthorCompanyExt(id=1325493871823692611, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, companyId=1325493871442010946, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Artificial Intelligence and Computer Science, Jiangnan University, Wuxi, Jiangsu 214122), AuthorCompanyExt(id=1325493871865635652, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, companyId=1325493871442010946, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=江南大学人工智能与计算机学院 江苏无锡 214122)])]), Author(id=1325493878534579026, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1325493878970786644, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, authorId=1325493878534579026, language=EN, stringName=Jianjun Liu, firstName=Jianjun, middleName=null, lastName=Liu, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=School of Artificial Intelligence and Computer Science, Jiangnan University, Wuxi, Jiangsu 214122, bio={"content":"

born in 1987. PhD, associate professor. His main research interests include artificial intelligence, computer vision, and pattern recognition

"}, bioImg=null, bioContent=

born in 1987. PhD, associate professor. His main research interests include artificial intelligence, computer vision, and pattern recognition

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325493881059550037, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, authorId=1325493878534579026, language=CN, stringName=刘建军, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=江南大学人工智能与计算机学院 江苏无锡 214122, bio={"img":"EsvLQpMB4mtaULFIEY6psg==","content":"

1987年生。博士,副教授。主要研究方向为人工智能、计算机视觉、模式识别

"}, bioImg=EsvLQpMB4mtaULFIEY6psg==, bioContent=

1987年生。博士,副教授。主要研究方向为人工智能、计算机视觉、模式识别

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325493871442010946, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, xref=null, ext=[AuthorCompanyExt(id=1325493871823692611, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, companyId=1325493871442010946, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Artificial Intelligence and Computer Science, Jiangnan University, Wuxi, Jiangsu 214122), AuthorCompanyExt(id=1325493871865635652, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325493828240679703, companyId=1325493871442010946, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=江南大学人工智能与计算机学院 江苏无锡 214122)])])]
张飞翔 , 李志华* , 刘建军
作者信息
  • 江南大学人工智能与计算机学院 江苏无锡 214122
通讯作者:
李志华()
作者简介:

1998年生。硕士研究生。主要研究方向为压缩感知、图像重建

1969年生。博士,教授。主要研究方向为人工智能及其安全、云计算与云数据中心理论、大数据挖掘、图像重建、边缘计算、信息安全

1987年生。博士,副教授。主要研究方向为人工智能、计算机视觉、模式识别

Stable Image Reconstruction Method Based on Resilient Total Variation Minimization
Feixiang Zhang , Zhihua Li* , Jianjun Liu
Affiliations
  • School of Artificial Intelligence and Computer Science, Jiangnan University, Wuxi, Jiangsu 214122
  • born in 1998. Master candidate. His main research interests include compressed sensing and image reconstruction

    born in 1969. PhD, professor. His main research interests include artificial intelligence and its security, cloud computing and cloud data center theory, big data mining, image reconstruction, edge computing, and information security

    born in 1987. PhD, associate professor. His main research interests include artificial intelligence, computer vision, and pattern recognition

出版时间: 2026-02-01 doi: 10.7544/issn1000-1239.202550014
文章导航
收藏切换

在图像重建领域,传统的全变分(total variation,TV)会导致重建图像出现阶梯效应和对比度损失。针对此问题,利用非凸罚函数对TV进行改进并引入后向扩散处理机制,基于此,提出一种基于弹性全变分的图像重建模型。首先,为了克服重建图像的阶梯效应,用变形${l_1}$非凸罚函数对TV进行改进,以此加强模型对约束等距性质(restricted isometry property,RIP)的鲁棒性。其次,为了降低重建图像的对比度损失并抑制噪声,通过在模型中引入后向扩散处理机制,以便保证图像均匀区域的平滑性。然后,基于RIP条件推导了模型稳定重建的理论保证。最后,联合DC(difference of convex functions)规划和交替方向乘子法对模型进行求解,基于此进一步提出一个基于弹性全变分的图像重建方法。为了验证所提出方法的有效性、高效性和实用性,分别使用标准图像、自然图像和医学图像进行重建实验。实验结果表明,与已有方法相比,所提出方法在欠采样场景和噪声环境中的重建质量均有不同程度的改善和提高。

压缩感知  /  图像重建  /  全变分  /  约束等距性质  /  后向扩散

In the field image reconstruction, traditional total variation (TV) regularization often incurs staircase artifacts and compromises image contrast. In this work, we refine the total variation framework by incorporating a non-convex penalty function alongside a backward diffusion mechanism; on this basis, we propose a resilient total variation (RTV)-based image reconstruction model. First, to mitigate staircase artifacts, the proposed method enhances the robustness of TV regularization by incorporating the transformed ${l_1}$ non-convex penalty functions, which improve the model’s adherence to the restricted isometry property (RIP). Second, to mitigate contrast loss in the reconstructed image and suppress noise, a backward diffusion mechanism is incorporated into the model to ensure smoothness in homogeneous regions. Subsequently, based on the RIP conditions, we derive theoretical guarantees for the stable reconstruction of the RTV image reconstruction model. The developed model is solved using a combination scheme of DC (difference of convex functions) programming and the alternating direction multiplier method, yielding an efficient resilient total variation-based reconstruction method. Extensive experiments on standard, natural, and medical images validate the method’s efficacy. The results demonstrate that the proposed method achieves superior reconstruction quality in comparison to the state-of-the-art methods, particularly in under-sampled and noisy scenarios, highlighting its effectiveness, efficiency, and practical utility.

compressed sensing  /  image reconstruction  /  total variation  /  restricted isometry property  /  backward diffusion
张飞翔, 李志华, 刘建军. 基于弹性全变分最小化的稳定图像重建方法. 计算机研究与发展, 2026 , 63 (2) : 460 -476 . DOI: 10.7544/issn1000-1239.202550014
Feixiang Zhang, Zhihua Li, Jianjun Liu. Stable Image Reconstruction Method Based on Resilient Total Variation Minimization[J]. Journal of Computer Research and Development, 2026 , 63 (2) : 460 -476 . DOI: 10.7544/issn1000-1239.202550014
压缩感知(compressed sensing,CS)利用信号的稀疏性,可以在极低的采样率下重建信号,被广泛应用于无线通信和医学成像等领域[1-3]。基于压缩感知的图像重建旨在从观测值${\boldsymbol{Y}} = {\boldsymbol{AX}} + {\boldsymbol{e}} \in {\mathbb{R}^M}$中重建出原始图像${\boldsymbol{X}} \in {\mathbb{C}^{N \times N}}$,其中${\boldsymbol{A}}:{\mathbb{C}^{N \times N}} \to {\mathbb{C}^M}$表示测量矩阵,${\boldsymbol{e}} \in {\mathbb{C}^M}$表示随机噪声,且满足$ {\left\| {\boldsymbol{e}} \right\|_2} \leqslant \varepsilon $,$\varepsilon $表示噪声水平。学术界从多个角度对图像重建的理论和方法进行了研究,如全变分(total variation,TV)方法[4]、字典学习[5]方法、正则化方法[6]、低秩矩阵分解[7]方法、深度学习方法[8]和迭代收缩阈值算法[9]等。其中基于TV最小化模型的图像重建由于其依据图像梯度域的稀疏性实现,所以在图像重建和图像去噪等应用中表现优异[4,10]。通常,TV最小化图像重建问题可建模成如式(1)所示[11]。
$ \mathop {\min }\limits_{{{\boldsymbol{X}}} \in {\mathbb{C}^{N \times N}}} {\left\| {\nabla {\boldsymbol{X}}} \right\|_1}\;{\text{ s}}{\text{.t}}{\text{. }}{\left\| {{\boldsymbol{AX}} - {\boldsymbol{Y}}} \right\|_2} \leqslant \varepsilon \;,$
其中$\nabla : {\mathbb{C}^{N \times N}} \to {\mathbb{C}^{N \times N \times 2}}$表示梯度算子,$ {\left\| {\nabla {{\boldsymbol{X}}}} \right\|_1} = \displaystyle\sum_{i,j} {|{{[\nabla {{\boldsymbol{X}}}]}_{i,j}}|} $表示TV半范数,即离散化图像${\boldsymbol{X}}$在梯度域中所有元素$ |{[\nabla {{\boldsymbol{X}}}]_{i,j}}| $的和。在图像重建中,TV半范数的物理意义在于量化图像中所有局部变化的总强度。TV半范数最小化通过最小化梯度幅值的总和,生成平坦区域与尖锐边缘相结合的重建结果,平坦区域的梯度接近0,而尖锐边缘的局部梯度则比较显著。虽然,在重建图像方面,TV最小化模型能够在保留区域边缘的同时具备一定的噪声抑制能力,并且TV最小化模型能从较少的观测值中重建出原始图像。但是,TV最小化模型对渐变区域的强制梯度稀疏性处理机制容易导致重建图像产生阶梯效应,引起重建图像的细节丢失,造成重建图像的块状伪影;TV最小化模型对低对比度区域的过渡平滑处理策略容易导致边缘处的对比度丢失,因而造成重建图像边缘附近模糊。
针对重建图像因丢失细节出现伪影和丢失边缘处对比度这2个不足,面向TV最小化模型,文献[12-20]分别针对性地进行了比较深入的研究。一方面,针对重建图像中出现的阶梯效应,文献[12-17]提出了多种改进方案。例如,文献[12-13]通过构建包含高阶导数的广义全变分(total generalized variation,TGV)对重建模型进行改进,这类模型能够逼近任意阶的分段多项式函数,从而有效减弱阶梯效应带来的负面影响;文献[14-15]通过引入平均曲率正则项对TV最小化模型进行改进,在减弱阶梯效应的同时,较好地保持了图像的角点;文献[16-17]提出使用欧拉弹性正则项来抑制阶梯效应,并且类似方法能有效保留图像的边缘。然而,这些方法尚存些许不足,文献[12-17]都涉及复杂的高阶导数计算,因而导致求解图像重建模型的计算复杂性变高。为此,文献[18-20]通过非凸罚函数来对TV最小化模型进行改造,提出相应的方法。实验结果表明,非凸罚函数的改进策略对解的稀疏性有促进作用,起到了改善重建图像阶梯效应的效果,提升了重建图像的质量。文献[12-20]同样对TV最小化模型进行了改进研究,但是缺乏对对比度损失所造成的重建图像边缘模糊问题的充分考虑;文献[21-26]从偏微分方程求解的角度对TV最小化模型进行改进。其中文献[21-23]通过采用冲击滤波方法来处理图像中的边缘模糊问题,这些方法起到了保证图像边缘的不连续性和图像均匀区域的平滑性作用;文献[24-26]通过向基于TV的偏微分方程中加入前后向(forward and backward,FAB)扩散处理机制,起到了增强图像对比度并抑制噪声干扰的作用。这些方法[21-26]都使用了后向扩散处理机制,被看作是经典的病态问题[27],对应模型的求解计算复杂。
另外,深度学习在图像重建领域的应用研究近年来进展显著[28-30]。基于U-Net架构的CNN模型通过跳跃连接实现了多尺度特征融合,在医学影像重建应用中展现出优异的局部细节恢复能力[28]。文献[29]通过联合优化数据保真项、TV项和网络参数,弥补了DIP(deep image prior)在高噪声干扰下图像重建的不足,实现了无监督学习的图像高质量复原。文献[30]通过生成器与鉴别器的对抗训练,推动了重建结果在纹理真实性和视觉一致性上的突破,在低光照增强和图像去模糊应用中效果颇佳。然而,基于深度学习的图像重建方法仍面临一些挑战。一方面,基于黑箱模型的图像重建结果可解释性有所欠缺,尤其在医疗诊断等高风险领域可能带来可信度的争议[28-30];另一方面,有监督学习的范式通常依赖精确的成对训练数据,而实际应用场景中获取高质量标注数据的成本高昂。
为了同时克服TV最小化模型中的阶梯效应和重建图像边缘模糊的问题,本文针对阶梯效应使用变形${l_1}$罚函数对TV模型进行改进,针对重建图像边缘模糊问题引入后向扩散处理机制,通过对二者进行综合提出一种新的、基于弹性全变分最小化的图像重建模型RTV(resilient total variation);另外,通过借助著名的DC(difference of convex functions)规划[31]、ADMM(alternating direction method of multipliers)[32]方法对上述优化模型进行离散化求解。基于此,进一步提出一种新的基于弹性全变分最小化的稳定图像重建DCA-RTV(difference of convex function algorithm-resilient total variation)算法。
本文的主要贡献有3点:
1)在稀疏采样图像重建过程中,针对重建图像出现的阶梯效应和对比度损失问题,提出一种基于弹性全变分最小化的图像重建模型;
2)基于约束等距性质,对存在噪声干扰时的图像重建建立稀疏采样图像重建的误差界;
3)通过对所提出的优化模型进行离散化求解,提出一种新的结合DC规划和ADMM方法的图像重建方法。
实验结果表明,与已有类似方法相比,基于弹性全变分的图像重建方法能有效抑制重建图像的阶梯效应和对比度损失问题,同时该方法具有一定的抗噪声干扰能力。
通常,TV的离散化形式可分为各向异性、各向同性,二者研究思路基本相似[33-34],因此本文在各向异性TV框架下进行讨论。各向异性TV可表示为
$ {\left\| {\boldsymbol{X}} \right\|_{{\text{T}}{{\text{V}}_{\text{a}}}}} = \displaystyle\sum_{i,j} {\left| {{\nabla _x}{\boldsymbol{X}}} \right| + \left| {{\nabla _y}{\boldsymbol{X}}} \right|} \;,$
其中${\nabla _x}{\boldsymbol{X}}$和${\nabla _y}{\boldsymbol{X}}$表示图像${\boldsymbol{X}}$在水平方向和垂直方向的梯度。值得注意的是,对于压缩感知而言,TV最小化模型通过利用${l_1}$范数起到了促进解的稀疏性作用[35]。然而,${l_1}$范数存在容易过度压缩信号中的高分量[36]的不足。为了克服这一问题,在此借助变形${l_1}$非凸罚函数的TV形式对TV模型进行改进。变形${l_1}$罚函数形如式(3)所示。
$ R_a^{{\text{TL1}}}({\boldsymbol{x}}): = \frac{{(\alpha + 1){{\left\| {\boldsymbol{x}} \right\|}_1}}}{{\alpha + {{\left\| {\boldsymbol{x}} \right\|}_1}}} = \displaystyle\sum_{i = 1}^n {\frac{{(\alpha + 1) | {{x_i}} |}}{{\alpha + | {{x_i}} |}}} \;,$
其中${\boldsymbol{x}}$表示信号,$ \alpha $表示非负参数。
文献[37-38]研究指出,无论测量矩阵是否满足受限等距性质(restricted isometry property,RIP)条件,变形${l_1}$罚函数都具有较好的稀疏信号重构性能。文献[19]通过构建变形TV正则化的图像重建模型,从理论上证明了变形${l_1}$罚函数改进TV模型的可行性和有效性,其中的非凸变体如式(4)所示。
$ {\left\| {\nabla {\boldsymbol{X}}} \right\|_{{\text{TL1}}}} = \displaystyle\sum_{i,j} {\frac{{(\alpha + 1)|{{[\nabla {\boldsymbol{X}}]}_{i,j}}|}}{{\alpha + |{{[\nabla {\boldsymbol{X}}]}_{i,j}}|}}} \;。$
受文献[19]启发,为了保持重建图像边缘的对比度和增强模型对噪声的抗干扰能力,本文通过在式(4)中引入后向扩散处理机制。这样处理的好处是:从偏微分方程的角度来看,通过对TV模型求导可获得一个欧拉-拉格朗日方程,并且再通过引入后向扩散项$ - \dfrac{\beta }{2}\left\| {\nabla {\boldsymbol{X}}} \right\|_2^2$,从而起到去除噪声和减弱重建图像的对比度损失问题的双重作用[20]。在此,基于上述策略所构建的弹性全变分如式(5)所示。
$ R_{\alpha ,\,\beta }^{{\text{RTV}}}({\boldsymbol{X}}): = {\left\| {\nabla {\boldsymbol{X}}} \right\|_{{\text{TL1}}}} - \frac{\beta }{2}\left\| {\nabla {\boldsymbol{X}}} \right\|_2^2\;,$
其中$\alpha $和$\beta $为弹性全变分的非负参数。$\alpha $负责控制惩罚函数的非凸性,$\alpha $越大,非凸惩罚函数对梯度幅值的惩罚越接近${l_0}$范数,稀疏性越强,反之则越接近${l_1}$范数;$\beta $调节后向扩散机制对边缘对比度、抑制噪声的贡献。在2个参数的共同作用下,使得模型在平衡稀疏性和保真性方面具有“弹性”。
进一步,基于式(5)提出一种新的基于弹性全变分最小化的图像重建模型,如式(6)所示。
$ \mathop {\min }\limits_{{{\boldsymbol{X}}} \in {\mathbb{C}^{N \times N}}} R_{\alpha ,\,\beta }^{{\text{RTV}}}({\boldsymbol{X}})\;{\text{ s}}{\text{.t}}{\text{. }}{\left\| { {\boldsymbol{AX}} - {\boldsymbol{Y}}} \right\|_2} \leqslant \varepsilon \;。$
根据压缩感知理论的指引,本节基于伪径向线采样策略[39]为式(6)建立图像重建保证,即从数学的角度推导并证明保证图像重建精度和稳定性的误差界,其中包括理论基础、图像信号的稳定性分析、图像重建误差分析。
本节简要介绍图像重建理论推导过程中所用到的数学引理。
受文献[39]的启发,测量矩阵${\boldsymbol{A}} \in {\mathbb{C}^{M \times N}}$的构造基于伪径向线采样策略。测量矩阵由伪径向采样掩膜${\boldsymbol{S}}$与2维离散傅里叶变换矩阵$\mathcal{F}$的Hadamard积生成,即${\boldsymbol{A}} = {\boldsymbol{S}} \circ \mathcal{F}$。采样掩膜的伪随机径向线分布采用与文献[39]相同的策略,采样率根据实验需求由采样线条数$L$调整。
根据文献[37],有以下关于变形${l_1}$罚函数的性质:
引理1.
1)对于变形${l_1}$罚函数,设常数$\alpha > 0$且$t \in \mathbb{R}$,有如下不等式成立:
2)设${x_i},{x_j} \in \mathbb{R}$,则${p_\alpha }(t)$满足如下不等式:
3)若$\left| t \right| \leqslant 1$,则$\left| t \right| \leqslant {p_\alpha }(t)$。
双变量Haar小波系统作为一种完备正交系统,是一种实用的数字图像稀疏近似方法[11]。为了简化Haar小波的稀疏性分析,在此给出以下引理:
引理2.
1)设均值为0的图像${\boldsymbol{X}} \in {\mathbb{C}^{N \times N}}$,${c_{(k)}}({\boldsymbol{X}})$表示图像${\boldsymbol{X}}$的双变量Haar小波系数中第$k$大的元素,或者是双变量Haar小波变换$\mathcal{H}{\boldsymbol{X}}$中第$k$大的元素。则$\forall k \geqslant 1$,$\tilde C > 0$,有如下不等式成立:
2)设$N = {2^n}$,$h_{j,k}^l$表示第$l$个Haar小波在索引$(j,k)$处的值。则最多有$6n$个双变量Haar小波不是常函数,即满足如下不等式:
3)对于任意的$j$,$k$,$l$,双变量Haar小波满足${\left\| {\nabla h_{j,k}^l} \right\|_1} \leqslant 8$。
根据文献[19],对于图像${\boldsymbol{X}}$,总存在一个常数$\beta > 0$,使得${\left\| {\nabla {{\boldsymbol{X}}_\beta }} \right\|_{{\text{TL1}}}} \leqslant {1/2}$,其中$ {{\boldsymbol{X}}_\beta } = {{\boldsymbol{X}}/ \beta } $。为便于表示,本文省略常数$\beta $,用$ {\boldsymbol{X}} $表示$ {{\boldsymbol{X}}_\beta } $。设$\hat {\boldsymbol{X}}$表示图像重建模型的最优解,令${\boldsymbol{D}} = \hat {\boldsymbol{X}} - {\boldsymbol{X}}$,则$\nabla {\boldsymbol{D}}$表示${\boldsymbol{D}}$的梯度。那么对于变形TV,满足如下性质:
引理3. 若${\left\| {\nabla {\boldsymbol{X}}} \right\|_{{\text{TL1}}}} \leqslant {1/2}$,则有${\left\| {\nabla {\boldsymbol{D}}} \right\|_{{\text{TL1}}}} \leqslant 1$成立。
根据文献[11],通过在矩阵$ {{\boldsymbol{\varLambda }}} $的底部和顶部分别连接一行零来考虑由矩阵$ {{\boldsymbol{\varLambda}} } $获得的矩阵${{{\boldsymbol{\varLambda}} }_0}$和${{{\boldsymbol{\varLambda}} }^0}$。即表示为如下式所示的增广矩阵:
对于一个线性算子$\mathcal{A}:{\mathbb{C}^{\left( {N - 1} \right) \times N}} \to {\mathbb{C}^m}$,其第$j$个分量由$ \langle {{{\boldsymbol{A}}_j},{\boldsymbol{X}}} \rangle $表示,则图像及其梯度可以通过类似的关系表示成引理4。
引理4. 设图像${\boldsymbol{X}} \in {\mathbb{C}^{N \times N}}$,矩阵${\boldsymbol{\varLambda}} \in {\mathbb{C}^{\left( {N - 1} \right) \times N}}$,则有
$ \left\{ \begin{aligned}&\left\langle {{\boldsymbol{\varLambda}} ,{{\boldsymbol{X}}_x}} \right\rangle = \left\langle {{{\boldsymbol{\varLambda}} ^0},{\boldsymbol{X}}} \right\rangle - \left\langle {{{\boldsymbol{\varLambda}} _0},{\boldsymbol{X}}} \right\rangle \;, \\&\left\langle {{\boldsymbol{\varLambda}} ,{\boldsymbol{X}}_y^{\text{T}}} \right\rangle = \left\langle {{{\boldsymbol{\varLambda}} ^0},{{\boldsymbol{X}}^{\text{T}}}} \right\rangle - \left\langle {{{\boldsymbol{\varLambda}} _0},{{\boldsymbol{X}}^{\text{T}}}} \right\rangle \;。 \end{aligned} \right. $
根据文献[11],强Sobolev不等式定义如引理5所示。
引理5. 设$\mathcal{B}:{\mathbb{C}^{N \times N}} \to {\mathbb{C}^m}$表示一个线性映射使得变换$ \mathcal{B}{\mathcal{H}^ * }:{\mathbb{C}^{N \times N}} \to {\mathbb{C}^m} $具有$2s + 1$阶的约束等距性质,并且约束等距常数$\delta < 1$,其中$\mathcal{H}:{\mathbb{C}^{N \times N}} \to {\mathbb{C}^{N \times N}}$是双变量Haar小波变换。若${\boldsymbol{D}} \in {\mathbb{C}^{N \times N}}$满足管约束${\left\| {\mathcal{B}{\boldsymbol{D}}} \right\|_2} \leqslant \varepsilon $,那么有式(8)成立。
$ {\left\| {\boldsymbol{D}} \right\|_2} \leqslant {C_2}\left[ {\left( {\frac{{{{\left\| {\nabla {\boldsymbol{D}}} \right\|}_1}}}{{\sqrt s }}} \right)\ln \left( {\frac{{{N^2}}}{s}} \right) + \varepsilon } \right]\;。$
图像信号的重建稳定性给出了保证图像信号在重建过程中的误差控制和重建质量的理论依据。对于一个满足RIP条件的线性变换$\mathcal{A}$,当图像信号${\boldsymbol{V}}$接近$\mathcal{A}$的零空间时,应当确保图像${\boldsymbol{V}}$满足特定的几何条件(如管约束和锥约束),以保证信号的重建精度和稳定性[19-20]。当图像满足特定管约束和锥约束时,图像${\boldsymbol{V}}$的${l_1}$范数和${l_2}$范数应满足的条件如命题1所述。
命题1. 设常数$\gamma \geqslant 1$,$k > 0$,${\eta _1} > 0$,${\eta _2} > 0$,$\mathcal{E} > 0$;$\mathcal{A}:{\mathbb{C}^{{n_1} \times {n_2}}} \to {\mathbb{C}^{\tilde m}}$是满足$2s$阶的RIP算子,其中${n_1}$,${n_2} $,$\tilde m \in {\mathbb{N}_ + } $,且满足${\delta _{2s}} < \dfrac{{3{\alpha ^2} - 2\alpha - 1}}{{5{\alpha ^2} + 2\alpha + 1}}(\alpha > 0)$;设$ \forall S $、图像${\boldsymbol{V}} \in {\mathbb{C}^{N \times N}}$,且$ \left| S \right| \leqslant k $,${\left\| {\boldsymbol{V}} \right\|_\infty } \leqslant 1$,满足管约束$ {\left\| {\mathcal{A}{\boldsymbol{V}}} \right\|_2} \leqslant \mathcal{E} $,并同时满足锥约束$ {P_\alpha }\left( {{{\boldsymbol{V}}_{{S^{\rm c}}}}} \right) \leqslant \gamma {P_\alpha }\left( {{{\boldsymbol{V}}_S}} \right) - \dfrac{{{\eta _1}}}{2}\left\| {\boldsymbol{V}} \right\|_2^2 +\sigma + {\eta _2}\left\langle {{E_1},{E_2}} \right\rangle $,其中${P_\alpha }({\boldsymbol{x}}) = \dfrac{{(\alpha + 1){{\left\| {\boldsymbol{x}} \right\|}_1}}}{{\alpha + {{\left\| {\boldsymbol{x}} \right\|}_1}}}$,${E_1}$,${E_2}$表示常数、向量或矩阵,并且规定${\left\| {{E_2}} \right\|_2} = {\left\| {\boldsymbol{V}} \right\|_2}$;进一步假设常数${\eta _2}$满足
$ {\eta _2} \leqslant \frac{{\gamma \sqrt k }}{{2{K_2}{{\left\| {{E_1}} \right\|}_2}}}\;,$
则图像的${l_2}$范数满足式(10)所示的关系。
$ {\left\| {\boldsymbol{V}} \right\|_2} \leqslant \sqrt {\frac{{\gamma \sqrt k {K_1}}}{{{\eta _1}{K_2}}}\mathcal{E} + \frac{2}{{{\eta _1}}}\sigma } \;,$
其中
$ {K_1}: = \frac{{3\alpha + 1}}{{2\alpha \sqrt {1 - {\delta _{2s}}} - \left( {1 + \alpha } \right)\sqrt {1 + {\delta _{2s}}} }}\;,$
$ \begin{split}{K_2}: =\;& \frac{{\sqrt {1 + {\delta _{2s}}} }}{4}\left( {{K_1} + \frac{1}{{\sqrt {1 + {\delta _{2s}}} }}} \right) = \\&\dfrac{{\sqrt {1 - {\delta _{2s}}} + \sqrt {1 + {\delta _{2s}}} }}{{4\sqrt {1 - {\delta _{2s}}} - 2\left( {\dfrac{{\alpha + 1}}{\alpha }} \right)\sqrt {1 + {\delta _{2s}}} }}\;。\end{split} $
图像的${l_1}$范数满足式(13)所示关系。
$ \begin{split}{\left\| {\boldsymbol{V}} \right\|_1} \leqslant \;&\frac{{\left( {2(\alpha + 1){K_2} + \alpha } \right)\gamma \sqrt k + 2(\alpha + 1){K_2}\sqrt k }}{{2\alpha {K_2}}}{\left\| {\boldsymbol{V}} \right\|_2} + \sigma \leqslant \\&\frac{{\left( {2(\alpha + 1){K_2} + \alpha } \right)\gamma \sqrt k + 2(\alpha + 1){K_2}\sqrt k }}{{2\alpha {K_2}}} \times \\&\sqrt {\frac{{\gamma \sqrt k {K_1}}}{{{\eta _1}{K_2}}}\mathcal{E} + \frac{2}{{{\eta _1}}}\sigma } + \sigma \;。 \\[-1pt]\end{split} $
针对命题1的证明,首先将${{\boldsymbol{V}}_{{S^{\rm c}}}}$划分为若干子集,并对每个子集的范数进行上界估计;然后,利用锥约束和管状约束进一步收紧$ {\boldsymbol{V}} $的范数上界,并引入常数简化不等式;最后,通过验证条件得到$ {\boldsymbol{V}} $的范数的最终上界,从而完成命题1的证明。具体证明过程如下:
证明. 把支撑集补集${S^{\text{c}}}$中的元素按绝对值从大到小降序排列得到集合${S^{{\text{cd}}}}$,再把集合${S^{{\text{cd}}}}$分成大小为$4k{\gamma ^2}$的子集,从而有
其中$r = \left\lfloor {\dfrac{{{N^2} - \left| S \right|}}{{4k{\gamma ^2}}}} \right\rfloor $。
那么${{\boldsymbol{V}}_{{S_1}}}$包含了${{\boldsymbol{V}}_{{S^{\text{c}}}}}$中较大的$4k{\gamma ^2}$个元素,${{\boldsymbol{V}}_{{S_2}}}$,${{\boldsymbol{V}}_{{S_3}}}$依次变小。显然,${{\boldsymbol{V}}_{{S_j}}}$中的每一个元素都小于${{\boldsymbol{V}}_{{S_{j - 1}}}}$中元素的平均值,基于此,有式(14)成立。
$ \left\| {{{\boldsymbol{V}}_{{S_j}}}} \right\|_2^2 \leqslant 4k{\gamma ^2}{\left( {\frac{{{{\left\| {{{\boldsymbol{V}}_{{S_{j - 1}}}}} \right\|}_1}}}{{4k{\gamma ^2}}}} \right)^2} = \frac{{\left\| {{{\boldsymbol{V}}_{{S_{j - 1}}}}} \right\|_1^2}}{{4k{\gamma ^2}}} \;。$
根据式(14),结合2.1.2节的引理1与锥约束可得式(15)成立。
$ \begin{split}\displaystyle\sum_{j = 2}^r &{{{\left\| {{{\boldsymbol{V}}_{{S_j}}}} \right\|}_2}} \leqslant \frac{1}{{2\gamma \sqrt k }}{\left\| {{{\boldsymbol{V}}_{{S^{\rm c}}}}} \right\|_1} \leqslant \frac{1}{{2\gamma \sqrt k }}{\left\| {{{\boldsymbol{V}}_{{S^{\text{c}}}}}} \right\|_{\rm TL1}} \leqslant \\& \frac{{{{\left\| {{{\boldsymbol{V}}_S}} \right\|}_{{\rm TL1}}}}}{{2\sqrt k }} - \frac{{{\eta _1}}}{{4\gamma \sqrt k }}\left\| {\boldsymbol{V}} \right\|_2^2 + \frac{\sigma }{{2\gamma \sqrt k }} + \frac{{{\eta _2}}}{{2\gamma \sqrt k }}\left\langle {{E_1},{E_2}} \right\rangle \leqslant \\&\left( {\frac{{\alpha + 1}}{\alpha }} \right)\frac{{{{\left\| {{{\boldsymbol{V}}_S} + {{\boldsymbol{V}}_{{S_1}}}} \right\|}_2}}}{2} - \frac{{{\eta _1}}}{{4\gamma \sqrt k }}\left\| {\boldsymbol{V }}\right\|_2^2 + \\&\frac{\sigma }{{2\gamma \sqrt k }} + \frac{{{\eta _2}}}{{2\gamma \sqrt k }}\left\langle {{E_1},{E_2}} \right\rangle \;。\\[-1pt]\end{split} $
根据管约束$ {\left\| {\mathcal{A}{\boldsymbol{V}}} \right\|_2} \leqslant \mathcal{E} $,结合式(5)以及$\mathcal{A}$的RIP条件得到式(16)。
$ \begin{split}\mathcal{E} \geqslant\;& {\left\| {\mathcal{A}{\boldsymbol{V}}} \right\|_2} \geqslant {\left\| {\mathcal{A}\left( {{{\boldsymbol{V}}_S} + {{\boldsymbol{V}}_1}} \right)} \right\|_2} - \displaystyle\sum_{j = 2}^r {{{\left\| {\mathcal{A}{{\boldsymbol{V}}_{{S_j}}}} \right\|}_2}} \geqslant \\&\sqrt {1 - {\delta _{2s}}} {\left\| {{{\boldsymbol{V}}_S} + {{\boldsymbol{V}}_1}} \right\|_2} - \sqrt {1 + {\delta _{2s}}} \displaystyle\sum_{j = 2}^r {{{\left\| {{{\boldsymbol{V}}_{{S_j}}}} \right\|}_2}} \geqslant \\&\sqrt {1 - {\delta _{2s}}} {\left\| {{{\boldsymbol{V}}_S} + {{\boldsymbol{V}}_1}} \right\|_2} - \sqrt {1 + {\delta _{2s}}} \Bigg( {\left( {1 + \frac{1}{\alpha }} \right)\frac{{{{\left\| {{{\boldsymbol{V}}_S} + {{\boldsymbol{V}}_{{S_1}}}} \right\|}_2}}}{2}} -\\ & {\frac{{{\eta _1}}}{{4\gamma \sqrt k }}\left\| {\boldsymbol{V}} \right\|_2^2 + \frac{\sigma }{{2\gamma \sqrt k }} + \frac{{{\eta _2}}}{{2\gamma \sqrt k }}\left\langle {{E_1},{E_2}} \right\rangle } \Bigg) \geqslant \Bigg( \sqrt {1 - {\delta _{2s}}} -\\&\frac{{\left( {1 + \alpha } \right)\sqrt {1 + {\delta _{2s}}} }}{{2\alpha }} \Bigg){\left\| {{{\boldsymbol{V}}_S} + {{\boldsymbol{V}}_1}} \right\|_2} +\frac{{{\eta _1}\sqrt {1 + {\delta _{2s}}} \left\| {\boldsymbol{V}} \right\|_2^2}}{{4\gamma \sqrt k }} -\\&\frac{{\sigma \sqrt {1 + {\delta _{2s}}} }}{{2\gamma \sqrt k }} - \frac{{{\eta _2}\sqrt {1 + {\delta _{2s}}} }}{{2\gamma \sqrt k }}\left\langle {{E_1},{E_2}} \right\rangle \;。\\[-1pt]\end{split} $
根据命题1,假设${\delta _{2s}} < \dfrac{{3{\alpha ^2} - 2\alpha - 1}}{{5{\alpha ^2} + 2\alpha + 1}}$,可知$ \sqrt {1 - {\delta _{2s}}} - \dfrac{{\left( {1 + \alpha } \right)\sqrt {1 + {\delta _{2s}}} }}{{2\alpha }} > 0 $。据此,进一步对式(16)进行化简可得式(17)成立。
$ \begin{split}&{\left\| {{{\boldsymbol{V}}_S} + {{\boldsymbol{V}}_{{S_1}}}} \right\|_2} \leqslant \frac{{2\alpha }}{{2\alpha \sqrt {1 - {\delta _{2s}}} - \left( {1 + \alpha } \right)\sqrt {1 + {\delta _{2s}}} }}\Bigg( {\varepsilon - } \\&\quad {\frac{{{\eta _1}\sqrt {1 + {\delta _{2s}}} \left\| {\boldsymbol{V}} \right\|_2^2}}{{4\gamma \sqrt k }} + \frac{{\sigma \sqrt {1 + {\delta _{2s}}} }}{{2\gamma \sqrt k }} + \frac{{{\eta _2}\sqrt {1 + {\delta _{2s}}} }}{{2\gamma \sqrt k }}\left\langle {{E_1},{E_2}} \right\rangle } \Bigg)\;。\end{split} $
由于$ {\left\| {\boldsymbol{V}} \right\|_2} \leqslant {\left\| {{{\boldsymbol{V}}_S} + {{\boldsymbol{V}}_{{S_1}}}} \right\|_2} + \displaystyle\sum_{j = 2}^r {{{\left\| {{{\boldsymbol{V}}_{{S_j}}}} \right\|}_2}} $,通过把式(15)、式(17)代入这一关系式可得式(18)成立。
$ \begin{split}{\left\| {\boldsymbol{V}} \right\|_2} \leqslant& \frac{{3\alpha + 1}}{{2\alpha }}{\left\| {{{\boldsymbol{V}}_S} + {{\boldsymbol{V}}_{{S_1}}}} \right\|_2} - \frac{{{\eta _1}}}{{4\gamma \sqrt k }}\left\| {\boldsymbol{V}} \right\|_2^2 + \frac{\sigma }{{2\gamma \sqrt k }} +\\& \frac{{{\eta _2}}}{{2\gamma \sqrt k }}\left\langle {{E_1},{E_2}} \right\rangle \leqslant \frac{{(3\alpha + 1)\mathcal{E}}}{{2\alpha \sqrt {1 - {\delta _{2s}}} - \left( {1 + \alpha } \right)\sqrt {1 + {\delta _{2s}}} }} + \\&\left( {\frac{{3\alpha + 1}}{{2\alpha \sqrt {1 - {\delta _{2s}}} - \left( {1 + \alpha } \right)\sqrt {1 + {\delta _{2s}}} }} + \frac{1}{{\sqrt {1 + {\delta _{2s}}} }}} \right) \times \\&\Bigg( { - \frac{{{\eta _1}\sqrt {1 + {\delta _{2s}}} \left\| {\boldsymbol{V}} \right\|_2^2}}{{4\gamma \sqrt k }} + \frac{{\sigma \sqrt {1 + {\delta _{2s}}} }}{{2\gamma \sqrt k }}} + \\&{\frac{{{\eta _2}\sqrt {1 + {\delta _{2s}}} }}{{2\gamma \sqrt k }}\left\langle {{E_1},{E_2}} \right\rangle } \Bigg): = {K_1}\mathcal{E} - \frac{{{\eta _1}{K_2}}}{{\gamma \sqrt k }}\left\| {\boldsymbol{V}} \right\|_2^2 +\\& \frac{{2{K_2}}}{{\gamma \sqrt k }}\sigma + \frac{{2{\eta _2}{K_2}}}{{\gamma \sqrt k }}\left\langle {{E_1},{E_2}} \right\rangle \;。\\[-1pt] \end{split} $
根据式(9)的假设,利用柯西-施瓦茨不等式得到$ {\left\| {\boldsymbol{V}} \right\|_2} - \dfrac{{2{\eta _2}{K_2}}}{{\gamma \sqrt k }}\,\left\langle {{E_1},\,{E_2}} \right\rangle \geqslant 0 $,结合式(18)可得式(19)成立:
$ \begin{split}\frac{{{\eta _1}{K_2}}}{{\gamma \sqrt k }}&\left\| {\boldsymbol{V}} \right\|_2^2 + {\left\| {\boldsymbol{V}} \right\|_2} - \frac{{2{\eta _2}{K_2}}}{{\gamma \sqrt k }}\left\langle {{E_1},{E_2}} \right\rangle - {K_1}\mathcal{E} - \\&\frac{{2{K_2}}}{{\gamma \sqrt k }}\sigma \leqslant \frac{{{\eta _1}{K_2}}}{{\gamma \sqrt k }}\left\| {\boldsymbol{V}} \right\|_2^2 - {K_1}\mathcal{E} - \frac{{2{K_2}}}{{\gamma \sqrt k }}\sigma \leqslant 0\;。 \end{split} $
通过求解$ \dfrac{{{\eta _1}{K_2}}}{{\gamma \sqrt k }}\left\| {\boldsymbol{V}} \right\|_2^2 - {K_1}\mathcal{E} - \dfrac{{2{K_2}}}{{\gamma \sqrt k }}\sigma \leqslant 0 $,可得命题1中的式(10)成立。
由于${\left\| {\boldsymbol{V}} \right\|_1} = {\left\| {{{\boldsymbol{V}}_S}} \right\|_1} + {\left\| {{{\boldsymbol{V}}_{{S^{\rm c}}}}} \right\|_1}$,结合锥约束条件可得
$ \begin{split}&{\left\| {\boldsymbol{V}} \right\|_1} \leqslant {\left\| {{{\boldsymbol{V}}_S}} \right\|_{{\text{TL1}}}} + {\left\| {{{\boldsymbol{V}}_{{S^{\text{c}}}}}} \right\|_{{\text{TL1}}}} \leqslant (\gamma + 1){\left\| {{{\boldsymbol{V}}_S}} \right\|_{{\text{TL1}}}} - \frac{{{\eta _1}}}{2}\left\| {\boldsymbol{V}} \right\|_2^2 + \sigma + \\&{\eta _2}\left\langle {{E_1},{E_2}} \right\rangle \leqslant \left( {\frac{{\left( {\alpha + 1} \right)\left( {\gamma + 1} \right)}}{\alpha }} \right)\sqrt k {\left\| {{{\boldsymbol{V}}_S}} \right\|_2} + \sigma + \\&\frac{{\gamma \sqrt k }}{{2{K_2}}}{\left\| {\boldsymbol{V}} \right\|_2} \leqslant \left( {\frac{{\left( {\alpha + 1} \right)\left( {\gamma + 1} \right)}}{\alpha }} \right)\sqrt k {\left\| {\boldsymbol{V}} \right\|_2} + \sigma + \frac{{\gamma \sqrt k }}{{2{K_2}}}{\left\| {\boldsymbol{V}} \right\|_2} = \\&\frac{{{{\left\| {\boldsymbol{V}} \right\|}_2}}}{{2\alpha {K_2}}}\left( {\left( {2(\alpha + 1){K_2} + \alpha } \right)\gamma \sqrt k + 2(\alpha + 1){K_2}\sqrt k } \right) + \sigma \;。\\[-1pt] \end{split} $
最后,将式(10)代入式(20)可推导式(13)成立。
证毕.
命题1通过式(10)和式(13)所示的条件提供了对信号大小和稀疏性量化的依据,以此保证重建的图像信号不会因噪声或采样不充分而变得不稳定。
图像重建误差是衡量图像重建算法稳定性的重要评价指标之一。对于基于伪径向线采样的图像重建,由于其采样模式在$k$空间中均匀分布,以下定理1给出了在伪径向线采样条件下图像重建模型(即式(6))的重建误差理论保证。
定理1. 设$N = {2^n}$,其中$n \in {\mathbb{N}_ + }$;设$ \mathcal{A}:{\mathbb{C}^{(N - 1) \times N}} \to {\mathbb{C}^{{M_1}}} $,$ \mathcal{A}':{\mathbb{C}^{(N - 1) \times N}} \to {\mathbb{C}^{{M_1}}} $这2个线性算子组成的合并算子$[\mathcal{A},\mathcal{A}']$满足$2s$阶的RIP条件;又设正交双变量Haar小波算子$\mathcal{H}:{\mathbb{C}^{N \times N}} \to {\mathbb{C}^{N \times N}}$与线性算子$ \mathcal{B}:{\mathbb{C}^{N \times N}} \to {\mathbb{C}^{{M_2}}} $构成复合算子$\mathcal{B}{\mathcal{H}^{ - 1}}:{\mathbb{C}^{N \times N}} \to {\mathbb{C}^{{M_2}}}$,并满足$2s + 1$阶的RIP条件;另设:$M = 4{M_1} + {M_2}$,以及线性算子$ {\varPhi} : {\mathbb{C}^{N \times N}} \to {\mathbb{C}^M}$,且满足式(21)。
$ \varPhi ({\boldsymbol{X)}} = \left( {{\mathcal{A}^0}({\boldsymbol{X}}),{\mathcal{A}_0}({\boldsymbol{X}}),\mathcal{A}{'^0}({\boldsymbol{X}}),{\mathcal{A}_0}({\boldsymbol{X}}),\mathcal{B}({\boldsymbol{X}})} \right) \;。$
设$ {\boldsymbol{X}} \in {\mathbb{C}^{N \times N}} $表示待重建图像,$\tilde {\boldsymbol{X}}$表示式(6)中模型最优解;进一步假设参数$\beta $满足
$ \beta\leqslant\frac{\sqrt{s}}{2K_2 \left\| \nabla\tilde{\boldsymbol{X}} \right\| _2}\; , $
则有稳定梯度重建结果式(23)(24)成立。
$ {\left\| {\nabla {\boldsymbol{X}} - \nabla \tilde {\boldsymbol{X}}} \right\|_2} \leqslant \sqrt {{\varsigma _1}\sqrt s \varepsilon + {\varsigma _2}{{\left\| {\nabla {\boldsymbol{X}} - {{\left( {\nabla {\boldsymbol{X}}} \right)}_S}} \right\|}_1}} \;,$
$ \begin{split}{\left\| {\nabla {\boldsymbol{X}} - \nabla \tilde {\boldsymbol{X}}} \right\|_1} \leqslant\; &{\varsigma _3}\sqrt s \sqrt {{\varsigma _1}\sqrt s \varepsilon + {\varsigma _2}{{\left\| {\nabla {\boldsymbol{X}} - {{\left( {\nabla {\boldsymbol{X}}} \right)}_S}} \right\|}_1}} + \\ &{\varsigma _4}{\left\| {\nabla {\boldsymbol{X}} - {{\left( {\nabla {\boldsymbol{X}}} \right)}_S}} \right\|_1}\;,\\[-1pt]\end{split} $
其中${\varsigma _1}$,${\varsigma _2}$,$ {\varsigma _3} $,$ {\varsigma _4} $是取决于${\delta _{2s}}$,$\alpha $,$\beta $的常数。
同时,有稳定图像重建结果式(25)成立。
$ \begin{split}{\left\| {{\boldsymbol{X}} - \tilde {\boldsymbol{X}}} \right\|_2} \lesssim\;& \ln \left( {\frac{{{N^2}}}{s}} \right)\sqrt {{\varsigma _1}\sqrt s \varepsilon + {\varsigma _2}{{\left\| {\nabla {\boldsymbol{X}} - {{\left( {\nabla {\boldsymbol{X}}} \right)}_S}} \right\|}_1}} + \\&\ln \left( {\frac{{{N^2}}}{s}} \right)\frac{{{{\left\| {\nabla {\boldsymbol{X}} - {{\left( {\nabla {\boldsymbol{X}}} \right)}_S}} \right\|}_1}}}{{\sqrt s }} + \varepsilon \;。\\[-1pt] \end{split} $
针对定理1的证明,策略性地将证明过程分为梯度稳定性重建和图像稳定性重建2部分。首先,对于梯度稳定性重建,根据命题1并结合锥约束和管状约束,对梯度误差进行上界估计;然后,对于图像稳定性重建,利用强Sobolev不等式[11]验证图像误差的稳定性,以此得到误差的最终上界。
证明. 令${P_\alpha }({\boldsymbol{x}}) = \dfrac{{(\alpha + 1){{\left\| {\boldsymbol{x}} \right\|}_1}}}{{\alpha + {{\left\| {\boldsymbol{x}} \right\|}_1}}}$,$S$表示$\nabla {\boldsymbol{X}}$中较大的$s$个元素的索引组成的集合,${\boldsymbol{G}} = \tilde {\boldsymbol{X}} - {\boldsymbol{X}}$,${\boldsymbol{L}} = \left( {{{\boldsymbol{G}}_x},{\boldsymbol{G}}_y^{\text{T}}} \right)$;根据2.1.4节的引理3,令${\left\| {\boldsymbol{G}} \right\|_\infty } \leqslant {1/2}$,则${\left\| {\nabla {\boldsymbol{G}}} \right\|_\infty } \leqslant 1$;$J$表示从$\nabla {\boldsymbol{G}}$中非零元素的索引到${\boldsymbol{L}}$中对应索引的映射。显然,${\boldsymbol{L}}$和$\nabla {\boldsymbol{G}}$具有相同的范数,即${\left\| {\boldsymbol{L}} \right\|_1} = {\left\| {\nabla {\boldsymbol{G}}} \right\|_1}$,${\left\| {\boldsymbol{L}} \right\|_2} = {\left\| {\nabla {\boldsymbol{G}}} \right\|_2}$,${P_\alpha }({\boldsymbol{L}}) = {P_\alpha }(\nabla {\boldsymbol{G}})$。由于$\tilde {\boldsymbol{X}}$表示式(6)模型的最优解,则有式(26)成立。
$ \begin{split}{P_\alpha }(\nabla \tilde {\boldsymbol{X}}) -& \frac{\beta }{2}\left\| {\nabla \tilde {\boldsymbol{X}}} \right\|_2^2 \leqslant {P_\alpha }(\nabla {\boldsymbol{X}}) - \frac{\beta }{2}\left\| {\nabla {\boldsymbol{X}}} \right\|_2^2 \leqslant\\&{P_\alpha }({(\nabla {\boldsymbol{X}})_S}) + {P_\alpha }({(\nabla {\boldsymbol{X}})_{{S^{\text{c}}}}}) - \frac{\beta }{2}\left\| {\nabla {\boldsymbol{X}}} \right\|_2^2\;,\end{split} $
其中式(26)的左边项满足
$ \begin{split}{P_\alpha }(\nabla \tilde {\boldsymbol{X}}) - &\frac{\beta }{2}\left\| {\nabla \tilde {\boldsymbol{X}}} \right\|_2^2 = {P_\alpha }({(\nabla {\boldsymbol{X}})_S} + {(\nabla {\boldsymbol{G}})_S}) +{P_\alpha }({(\nabla {\boldsymbol{X}})_{{S^{\text{c}}}}} + \\& {(\nabla {\boldsymbol{G}})_{{S^{\text{c}}}}}) - \frac{\beta }{2}\left\| {\nabla \tilde {\boldsymbol{X}}} \right\|_2^2 \geqslant {P_\alpha }({(\nabla {\boldsymbol{X}})_S}) -{P_\alpha }({(\nabla {\boldsymbol{G}})_S}) + \\&{P_\alpha }({(\nabla {\boldsymbol{X}})_{{S^{\rm c}}}}) - {P_\alpha }({(\nabla {\boldsymbol{G}})_{{S^{\rm c}}}}) - \frac{\beta }{2}\left( {\left\| {\nabla {\boldsymbol{X}}} \right\|_2^2 + } \right. \\&\left. {2\left\langle {\nabla {\boldsymbol{X}},\nabla {\boldsymbol{G}}} \right\rangle + \left\| {\nabla {\boldsymbol{G}}} \right\|_2^2} \right)\;。\\[-1pt] \end{split} $
对式(27)进行化简可得到式(28)成立。
$ \begin{split}{P_\alpha }({(\nabla {\boldsymbol{G}})_{{S^{\text{c}}}}}) \leqslant\;& {P_\alpha }({(\nabla {\boldsymbol{G}})_S}) + 2{P_\alpha }(\nabla {\boldsymbol{X}} - {(\nabla {\boldsymbol{X}})_S}) - \\&\frac{\beta }{2}\left\| {\nabla {\boldsymbol{G}}} \right\|_2^2 + \beta \left\langle {\nabla \tilde {\boldsymbol{X}},\nabla {\boldsymbol{G}}} \right\rangle \;。\end{split} $
一方面,由于${\boldsymbol{L}}$包含了$\nabla {\boldsymbol{G}}$中相同的非零元素,所以根据式(28)可知式(29)成立。
$ \begin{split}{P_\alpha }({{\boldsymbol{L}}_{J{{(S)}^{\text{c}}}}}) \leqslant\;& {P_\alpha }({{\boldsymbol{L}}_{J(S)}}) + 2{P_\alpha }(\nabla {\boldsymbol{X}} - {(\nabla {\boldsymbol{X}})_S}) - \\&\frac{\beta }{2}\left\| {\boldsymbol{L}} \right\|_2^2 + \beta \left\langle {\nabla \tilde {\boldsymbol{X}},\nabla {\boldsymbol{G}}} \right\rangle \;。 \end{split} $
根据命题1,可见式(29)本质上是一个锥约束。
另一方面,由于${\boldsymbol{G}}$满足式(30)所示的管约束:
$ \begin{split}\left\| {\varPhi {\boldsymbol{G}}} \right\|_2^2 =\;& \left\| {(\varPhi \tilde {\boldsymbol{X}} - {\boldsymbol{Y}}) - (\varPhi {\boldsymbol{X}} - {\boldsymbol{Y}})} \right\|_2^2 \leqslant 2\left\| {\varPhi \tilde {\boldsymbol{X}} - {\boldsymbol{Y}}} \right\|_2^2 + \\&2\left\| {\varPhi {\boldsymbol{X}} - {\boldsymbol{Y}}} \right\|_2^2 \leqslant 4{\varepsilon ^2}\;。\\[-1pt]\end{split} $
根据引理4的式(7)可知式(31)成立。
$ \left\{ \begin{aligned}&{\left| {\left\langle {{{\boldsymbol{A}}_j},{{\boldsymbol{G}}_x}} \right\rangle } \right|^2} = {\left| {\left\langle {{{[{{\boldsymbol{A}}_j}]}^0},{\boldsymbol{G}}} \right\rangle - \left\langle {{{[{{\boldsymbol{A}}_j}]}_0},{\boldsymbol{G}}} \right\rangle } \right|^2} \leqslant\\ & \quad\quad\quad\quad\quad\;\; 2{\left| {\left\langle {{{[{{\boldsymbol{A}}_j}]}^0},{\boldsymbol{G}}} \right\rangle } \right|^2} + 2{\left| {\left\langle {{{[{{\boldsymbol{A}}_j}]}_0},{\boldsymbol{G}}} \right\rangle } \right|^2}\;, \\ &{\left| {\left\langle {{\boldsymbol{A}'_j} ,{\boldsymbol{G}}_y^{\text{T}}} \right\rangle } \right|^2} = {\left| {\left\langle {{{[{\boldsymbol{A}'_j} ]}^0},{{\boldsymbol{G}}^{\text{T}}}} \right\rangle - \left\langle {{{[{\boldsymbol{A}'_j} ]}_0},{{\boldsymbol{G}}^{\text{T}}}} \right\rangle } \right|^2} \leqslant\\ & \quad\quad\quad\quad\quad\;\; 2{\left| {\left\langle {{{[{\boldsymbol{A}'_j} ]}^0},{{\boldsymbol{G}}^{\text{T}}}} \right\rangle } \right|^2} + 2{\left| {\left\langle {{{[{\boldsymbol{A}'_j} ]}_0},{{\boldsymbol{G}}^{\text{T}}}} \right\rangle } \right|^2}\;。 \end{aligned} \right. $
由式(31)进一步可知式(32)成立,
$ \begin{split}\left| {\left[ {\mathcal{A}\mathcal{A}'} \right]{\boldsymbol{L}}} \right|_2^2 =\;& \displaystyle\sum_{j = 1}^M {{{\left| {\left\langle {{{\boldsymbol{A}}_j},{{\boldsymbol{G}}_x}} \right\rangle } \right|}^2} + {{\left| {\left\langle {{\boldsymbol{A}'_j} ,{\boldsymbol{G}}_y^{\text{T}}} \right\rangle } \right|}^2}}\leqslant \\ &2\left\| {\varPhi ({\boldsymbol{G}})} \right\|_2^2 \leqslant 8{\varepsilon ^2}\;。\end{split} $
结合2.2节中的命题1,令$\gamma = 1$,$k = s$,${\eta _1} = {\eta _2} = \beta $,$ \sigma = 2{P_\alpha }(\nabla {\boldsymbol{X}} - {(\nabla {\boldsymbol{X}})_S}) $,$\mathcal{E} = 2\sqrt 2 \varepsilon $,$ {E_1} = \nabla \tilde {\boldsymbol{X}} $,$ {E_2} = \nabla {\boldsymbol{G}} $,由假设条件式(22)并结合式(10)推导可知式(33)成立。
$ \begin{split}\| {\nabla \tilde {\boldsymbol{X}} - \nabla {\boldsymbol{X}}} \|_2 = &{\left\| {\boldsymbol{L}} \right\|_2} \leqslant \sqrt {\frac{{2\sqrt 2 {K_1}\sqrt s }}{{\beta {K_2}}}\varepsilon + \frac{4}{\beta }{P_\alpha }\left( {\nabla {\boldsymbol{X}} - {{\left( {\nabla {\boldsymbol{X}}} \right)}_S}} \right)}: = \\ &\sqrt {{\varsigma _1}\sqrt s \varepsilon + {\varsigma _2}{{\left\| {\nabla {\boldsymbol{X}} - {{\left( {\nabla {\boldsymbol{X}}} \right)}_S}} \right\|}_1}} \;, \\[-1pt] \end{split} $
其中$ {\varsigma _1} = \dfrac{{2\sqrt 2 {K_1}}}{{\beta {K_2}}} = \dfrac{{4\sqrt 2 \left( {3\alpha + 1} \right)}}{{\alpha \beta \left( {\sqrt {1 - {\delta _{2s}}} + \sqrt {1 + {\delta _{2s}}} } \right)}} $,$ {\varsigma _2} = \dfrac{{4\left( {\alpha + 1} \right)}}{{\alpha \beta }} $。
另一方面,根据式(13)可得到式(34)成立。
$ \begin{split}{\left\| {\nabla \tilde {\boldsymbol{X}} - \nabla {\boldsymbol{X}}} \right\|_1} \leqslant\;& \frac{{\left( {2(\alpha + 1){K_2} + \alpha } \right)\gamma \sqrt k + 2(\alpha + 1){K_2}\sqrt k }}{{2\alpha {K_2}}} + \\ &\sqrt {\frac{{2\sqrt 2 {K_1}\sqrt s }}{{\beta {K_2}}}\varepsilon + \frac{4}{\beta }{P_\alpha }\left( {\nabla {\boldsymbol{X}} - {{\left( {\nabla {\boldsymbol{X}}} \right)}_S}} \right)} + \\ &2{P_\alpha }\left( {\nabla {\boldsymbol{X}} - {{\left( {\nabla {\boldsymbol{X}}} \right)}_S}} \right): = \\ & {\varsigma _3}\sqrt s \sqrt {{\varsigma _1}\sqrt s \varepsilon + {\varsigma _2}{{\left\| {\nabla {\boldsymbol{X}} - {{\left( {\nabla {\boldsymbol{X}}} \right)}_S}} \right\|}_1}} + \\ &{\varsigma _4}{\left\| {\nabla {\boldsymbol{X}} - {{\left( {\nabla {\boldsymbol{X}}} \right)}_S}} \right\|_1}\;, \\[-1pt] \end{split} $
其中$ {\varsigma _3} = \dfrac{{2\left( {2\alpha + 1} \right) \sqrt {1 - {\delta _{2s}}} + \left( {\alpha + 1} \right) \sqrt {1 + {\delta _{2s}}} }}{{\alpha \left( {\sqrt {1 - {\delta _{2s}}} + \sqrt {1 + {\delta _{2s}}} } \right)}} $,$ {\varsigma _4} = \dfrac{{2\left( {\alpha + 1} \right)}}{\alpha } $。
根据引理5,把式(34)代入式(8),可得到式(35)成立。
$ \begin{split}{\left\| {{\boldsymbol{X}} - \tilde {\boldsymbol{X}}} \right\|_2}\lesssim \;& \ln \left( {\frac{{{N^2}}}{s}} \right)\left( {\frac{{{{\left\| {\nabla \tilde {\boldsymbol{X}} - \nabla {\boldsymbol{X}}} \right\|}_1}}}{{\sqrt s }}} \right) + \varepsilon : = \\ &{\varsigma _3}\ln \left( {\frac{{{N^2}}}{s}} \right)\sqrt {{\varsigma _1}\sqrt s \varepsilon + {\varsigma _2}{{\left\| {\nabla {\boldsymbol{X}} - {{\left( {\nabla {\boldsymbol{X}}} \right)}_S}} \right\|}_1}} + \\ &{\varsigma _4}\ln \left( {\frac{{{N^2}}}{s}} \right)\frac{{{{\left\| {\nabla {\boldsymbol{X}} - {{\left( {\nabla {\boldsymbol{X}}} \right)}_S}} \right\|}_1}}}{{\sqrt s }} + \varepsilon \lesssim \\ & \ln \left( {\frac{{{N^2}}}{s}} \right)\sqrt {{\varsigma _1}\sqrt s \varepsilon + {\varsigma _2}{{\left\| {\nabla {\boldsymbol{X}} - {{\left( {\nabla {\boldsymbol{X}}} \right)}_S}} \right\|}_1}} + \\ &\ln \left( {\frac{{{N^2}}}{s}} \right)\frac{{{{\left\| {\nabla {\boldsymbol{X}} - {{\left( {\nabla {\boldsymbol{X}}} \right)}_S}} \right\|}_1}}}{{\sqrt s }} + \varepsilon \;。\\[-1pt]\end{split} $
证毕.
本质上,式(33)和式(34)提供了梯度重建的保证,确保图像的边缘、纹理等细节能够被准确恢复;式(35)给出了图像整体重建的误差界,旨在保证图像重建的整体准确性。
著名的DC规划作为一种分解非凸问题的数学优化工具,为非凸优化问题的求解提供了一个可行且高效的方案[31,40]。首先,本文借助DC规划对式(6)中的目标函数进行分解,从而把式(6)的优化问题转换为凸优化子问题;其次,重写上述凸优化子问题为增广拉格朗日函数的形式;然后,借助交替方向乘子法(alternating direction method of multipliers,ADMM)[32]设计子问题求解算法ADMM-Lagrange用于求解上述重写的增广拉格朗日函数;最后,对上述求解过程进行概括,并基于此,提出一种新的基于RTV的图像重建方法。
为了求解目标函数(即式(6)),首先对其进行DC分解。设${\mathcal{D}_x}{\boldsymbol{X}}$和$ {\mathcal{D}_y}{\boldsymbol{X}} $分别表示$\nabla {\boldsymbol{X}}$的水平和垂直分量,并把${\mathcal{D}_x}$和$ {\mathcal{D}_y} $视为2个算子;由于DC规划将式(6)所示的非凸函数分解为2个凸函数的差,并在迭代过程中利用后一个凸函数的次梯度进行线性化,故使用线性上界$\left\langle {{\boldsymbol{X}} - {{\boldsymbol{X}}^k},{\boldsymbol \xi ^k}} \right\rangle $重写式(6),得${\boldsymbol{X}}$的更新公式如式(36)所示。
$ \begin{split} {{\boldsymbol{X}}^{k + 1}}& \in\mathop {\arg\,\min }\limits_{{\boldsymbol{X}} \in {\mathbb{C}^{N \times N}}} \Bigg\{ \frac{{\alpha + 1}}{\alpha }{\left\| {{\mathcal{D}_x}{\boldsymbol{X}}} \right\|_1} + \frac{{\alpha + 1}}{\alpha }{\left\| {{\mathcal{D}_y}{\boldsymbol{X}}} \right\|_1} - \\& \left\langle {{\mathcal{D}_x}{\boldsymbol{X}},{\boldsymbol \xi ^k}} \right\rangle - \left\langle {{\mathcal{D}_y}{\boldsymbol{X}},{\boldsymbol \xi ^k}} \right\rangle {\text{s}}{\text{.t}}{\text{. }}\;{\left\| {{\boldsymbol{AX}} - {\boldsymbol{Y}}} \right\|_2} \leqslant \varepsilon \Bigg\} \;,\\[-1pt]\end{split}$
其中次梯度变量$ {\boldsymbol \xi ^k} $的计算式如式(37)所示。
$ {\boldsymbol \xi ^k} \in \partial \left( {\frac{{\alpha + 1}}{\alpha }{{\left\| {\nabla {\boldsymbol{X}}} \right\|}_1} - \frac{{\left( {\alpha + 1} \right){{\left\| {\nabla {\boldsymbol{X}}} \right\|}_1}}}{{\alpha + {{\left\| {\nabla {\boldsymbol{X}}} \right\|}_1}}} + \frac{\beta }{2}\left\| {\nabla {\boldsymbol{X}}} \right\|_2^2} \right) \;。$
为了对如式(36)所示的${l_1}$子问题进行求解,首先对式(36)进行重写,得到如式(38)所示的约束优化问题。
$ \begin{aligned}& \mathop {\min }\limits_{{\boldsymbol{X}} \in {\mathbb{C}^{N \times N}}} \left\{ {\frac{{\alpha + 1}}{\alpha }{{\left\| {{{\boldsymbol{d}}_x}} \right\|}_1} + \frac{{\alpha + 1}}{\alpha }{{\left\| {{{\boldsymbol{d}}_y}} \right\|}_1} - \left\langle {{{\boldsymbol{d}}_x},{\boldsymbol \xi ^k}} \right\rangle - \left\langle {{\boldsymbol d_y},{\boldsymbol \xi ^k}} \right\rangle } \right\}\;, \\& {\text{ s}}{\text{.t}}{\text{. }} \begin{aligned} &{\boldsymbol{AX}} - {\boldsymbol{Y}} - {\boldsymbol{z}} = 0\;, \\& {\boldsymbol{z}} \in \mathcal{B}(0,\tau ): = \left\{ {{\boldsymbol{x}} \in {\mathbb{R}^m}:{{\left\| {\boldsymbol{x}} \right\|}_2} \leqslant \varepsilon } \right\}\;, \\& {\mathcal{D}_x}{\boldsymbol{X}} = {{\boldsymbol{d}}_x},{\mathcal{D}_y}{\boldsymbol{X}} = {{\boldsymbol{d}}_y}\;。\end{aligned}\end{aligned} $
对于式(38),通过引入拉格朗日乘子${\boldsymbol{\lambda}} $,${{\boldsymbol{b}}_x}$,${{\boldsymbol{b}}_y}$对式(38)进行去约束操作,得到增广拉格朗日函数如式(39)所示。
$ \begin{split}\mathcal{L}&\left( {{\boldsymbol{X}},{{\boldsymbol{d}}_x},{{\boldsymbol{d}}_y},{\boldsymbol{z}},{{\boldsymbol{b}}_x},{{\boldsymbol{b}}_y},{\boldsymbol{\lambda}} } \right): = \frac{{\alpha + 1}}{\alpha }{\left\| {{{\boldsymbol{d}}_x}} \right\|_1} + \frac{{\alpha + 1}}{\alpha }{\left\| {{{\boldsymbol{d}}_y}} \right\|_1} - \\&\left\langle {{{\boldsymbol{d}}_x},{\boldsymbol \xi ^k}} \right\rangle - \left\langle {{{\boldsymbol{d}}_y},{\boldsymbol \xi ^k}} \right\rangle + \frac{\mu }{2}\left\| {{\boldsymbol{z}} - \left( {{\boldsymbol{AX}} - {\boldsymbol{Y}}} \right) - {\boldsymbol{\lambda}} } \right\|_2^2 +\\&\frac{\sigma }{2}\left\| {{{\boldsymbol{d}}_x} - {\mathcal{D}_x}{\boldsymbol{X}} - {{\boldsymbol{b}}_x}} \right\|_2^2 + \frac{\sigma }{2}\left\| {{{\boldsymbol{d}}_y} - {\mathcal{D}_y}{\boldsymbol{X}} - {{\boldsymbol{b}}_y}} \right\|_2^2\;,\\[-1pt]\end{split} $
其中$\mu $和$\sigma $是非负惩罚参数。
这样,式(6)的求解就转化为一个无约束的优化问题求解。
为了求解式(39)所示的子问题,本文使用著名的ADMM[32]算法。ADMM方法主要通过将问题分解为几个可独立求解的子问题的策略,从而简化求解过程,并使求解过程更高效。具体过程如下:
首先,设${\boldsymbol{u}}$表示迭代过程中对当前图像${\boldsymbol{X}}$的估计,通过对${\boldsymbol{u}}$的2个分量求偏导,得到变量${\boldsymbol{u}}$的更新公式如式(40)所示。
$ \begin{split} {\boldsymbol{u}} = &{\left( {\mu {{\boldsymbol{A}}^{\text{T}}}{\boldsymbol{A}} + \sigma {\nabla ^{\text{T}}}\nabla } \right)^{ - 1}}\big( \mu {{\boldsymbol{A}}^{\text{T}}}\left( {{\boldsymbol{Y}} + {\boldsymbol{z}} - {\boldsymbol{\lambda}} } \right) + \\&\sigma \mathcal{D}_x^{\text{T}}({{\boldsymbol{d}}_x} - {{\boldsymbol{b}}_x}) + \sigma \mathcal{D}_y^{\text{T}}({{\boldsymbol{d}}_y} - {{\boldsymbol{b}}_y}) \big)\;。\end{split}$
其次,由于变量${{\boldsymbol{d}}_x}$和${{\boldsymbol{d}}_y}$在式(39)中的求解是关于${l_1}$正则化的二次问题,故使用软阈值算子[41]更新变量。更新表达式如式(41)(42)所示。
$ {{\boldsymbol{d}}_x} = shrink\left( {{\mathcal{D}_x}{\boldsymbol{u}} + {{\boldsymbol{b}}_x} + \frac{{\boldsymbol \xi _x^k}}{\sigma },\frac{{\alpha + 1}}{{\alpha \sigma }}} \right) \;,$
$ {{\boldsymbol{d}}_y} = shrink\left( {{\mathcal{D}_y}{\boldsymbol{u}} + {{\boldsymbol{b}}_y} + \frac{{\boldsymbol \xi _y^k}}{\sigma },\frac{{\alpha + 1}}{{\alpha \sigma }}} \right) \;。$
然后,因为式(39)中变量$\boldsymbol z$的更新主要是为了保证其满足式(36)中的约束条件${\left\| {{\boldsymbol{AX}} - {\boldsymbol{Y}}} \right\|_2} \leqslant \varepsilon $,为此,在此使用文献[20]中的投影算子$ {\mathcal{P}_{\mathcal{B}\left( {0,\tau } \right)}} $对变量$\boldsymbol z$进行更新。更新公式如式(43)所示。
$ {\boldsymbol{z}} = {\mathcal{P}_{\mathcal{B}\left( {0,\tau } \right)}}\left( {{\boldsymbol{Au}} - {\boldsymbol{Y}} + {\boldsymbol{\lambda}} } \right) \;。$
最后,使用对偶更新步骤更新式(39)中的拉格朗日乘子${\boldsymbol{\lambda}} $,$ {{\boldsymbol{b}}_x} $,${{\boldsymbol{b}}_y}$,更新过程分别如式(44)~(46)所示。
$ {{\boldsymbol{b}}_x} = {{\boldsymbol{b}}_x} + {\mathcal{D}_x}{\boldsymbol{u}} - {{\boldsymbol{d}}_x} \;,$
$ {{\boldsymbol{b}}_y} = {{\boldsymbol{b}}_y} + {\mathcal{D}_y}{\boldsymbol{u}} - {{\boldsymbol{d}}_y} \;,$
$ {\boldsymbol{\lambda}} = {\boldsymbol{\lambda}} + \left( {{\boldsymbol{Au}} - {\boldsymbol{Y}}} \right) - {\boldsymbol{z}} \;。$
概括上述求解过程,提出一种对子问题求解的ADMM-Lagrange算法,其伪代码描述如算法1所示。
算法1. ADMM-Lagrange算法。
输入:测量矩阵${\boldsymbol{A}}$,测量值${\boldsymbol{Y}}$,线性化变量$ {\boldsymbol \xi ^k} $;
输出:对目标图像的估计值${\boldsymbol{u}}$。
①初始化图像${\boldsymbol{X}} = {{0}}$,拉格朗日乘子${{\boldsymbol{b}}_x} = {{\boldsymbol{b}}_y} = $ ${\boldsymbol{\lambda}} = {{0}} $,梯度分量${{\boldsymbol{d}}_x} = {{\boldsymbol{d}}_y} = {{0}}$,循环变量$p = 0$, 最大迭代次数$ MaxADMM $;
② while $p < MaxADMM$ do
③  根据式(40)计算${\boldsymbol{u}}$;
④  根据式(41)计算${{\boldsymbol{d}}_x}$;
⑤  根据式(42)计算${{\boldsymbol{d}}_y}$;
⑥  根据式(43)计算${\boldsymbol{z}}$;
⑦  根据式(44)计算${{\boldsymbol{b}}_x}$;
⑧  根据式(45)计算${{\boldsymbol{b}}_y}$;
⑨  根据式(46)计算${\boldsymbol{\lambda}} $;
⑩   $p \leftarrow p + 1$;
⑪ end while
⑫ return ${\boldsymbol{u}}$。
综上,通过所提出的ADMM-Lagrange算法对式(36)中的图像${\boldsymbol{X}}$进行迭代逼近,也即${{\boldsymbol{X}}^k} = ADMM - Lagrange({\boldsymbol{Y}},{\boldsymbol \xi ^k})$。基于此,根据DC规划进一步提出RTV方法,用于求解式(6)所示的优化问题。RTV方法首先对式(6)进行DC分解,然后调用算法1根据测量值${\boldsymbol{Y}}$迭代逼近待重建图像${\boldsymbol{X}}$,最后返回迭代过程中的重建图像最优解$\tilde {\boldsymbol{X}}$。RTV方法的伪代码描述如算法2所示。
算法2. RTV算法。
输入:测量矩阵${\boldsymbol{A}}$,测量值${\boldsymbol{Y}}$;
输出:对目标图像的估计值$\tilde {\boldsymbol{X}}$。
① 初始化图像${{\boldsymbol{X}}^0} = {{0}}$,循环变量$k = 0$,最大迭代 次数$ MaxDCA $和容差阈值$ tol $;
② while $k < MaxDCA$ and ${\left\| {{{\boldsymbol{X}}^k} - {{\boldsymbol{X}}^{k - 1}}} \right\|_2} > tol$ do
③  根据式(37)计算线性化变量$ {\boldsymbol \xi ^k} $;
④   ${{\boldsymbol{X}}^k} = ADMM - Lagrange({\boldsymbol{Y}},{\boldsymbol \xi ^k})$;/*调用算法1 求解子问题(如式(36)所示)*/
⑤   $k \leftarrow k + 1$;
⑥   $\tilde {\boldsymbol{X}} = {{\boldsymbol{X}}^k}$;
⑦ end while
⑧ return $\tilde {\boldsymbol{X}}$。
本节将所提出的RTV方法应用于医学领域的核磁共振图像重建实验,并与其他经典的全变分方法TV[11],TVa-TVi[18],ETV[20]进行比较,以验证RTV方法的有效性、高效性和实用性。
本文同样采用业界公认的评价指标,包括:结构相似性指数测定(structural similarity index measure,SSIM)[42]、峰值信噪比(peak signal to noise ratio,PSNR)[19]、均方根误差(root mean square error,$ RMS E $)[12]和相对误差(relative error,$ RE $)[20]作为评价指标。
SSIM如式(47)所示。$ S S IM $综合考虑了图像的亮度、对比度和结构,是衡量图像质量或相似性的重要评价指标。$ S S IM $的值越接近1表明所对比的图像在视觉上越相似。
$ S S IM = \frac{{(2{\mu _x}{\mu _y} + {C_1})(2{\sigma _{xy}} + {C_2})}}{{(\mu _x^2 + \mu _y^2 + {C_1})(\sigma _x^2 + \sigma _y^2 + {C_2})}}\;,$
其中${\mu _x}$与${\mu _y}$为2幅图像的均值,$\sigma _x^2$与$\sigma _y^2$分别为2幅图像的方差,${\sigma _{xy}}$为2幅图像的协方差,${C_1}$与${C_2}$是用来维持稳定的常数。
PSNR计算公式如式(48)所示。$ PS NR $用于衡量2个图像噪声水平的差异,$ PS NR $的值越高,表明重建图像的质量越好。
$ PS NR = 10 \times \lg \left( {\dfrac{{255}}{{M \times N \displaystyle\sum_{i = 1}^N {\displaystyle\sum_{j = 1}^N {{{(\tilde {\boldsymbol{X}}(i,j) - {\boldsymbol{X}}(i,j))}^2}} } }}} \right)\;,$
其中$\tilde {\boldsymbol{X}}$表示重建图像,${\boldsymbol{X}}$表示原图像,$M$与$N$表示图像的尺寸大小。
RMSE如式(49)所示。$ RMS E $量化了重建图像与原始图像之间的平均误差大小。
$ RMS E = \sqrt {\frac{1}{{M \times N}}\displaystyle\sum_{i = 1}^N {\displaystyle\sum_{j = 1}^N {{{(\tilde {\boldsymbol{X}}(i,j) - {\boldsymbol{X}}(i,j))}^2}} } } \;,$
其中变量$\tilde {\boldsymbol{X}}$,${\boldsymbol{X}}$,$M$,$N$的含义与式(48)中的一致。
RE如式(50)所示。$ RE $计算重建图像与原始图像之间的误差相对于原始图像的比例,主要用于评价重建过程中误差的相对大小。
$ RE = \frac{{{{\left\| {\tilde {\boldsymbol{X}} - {\boldsymbol{X}}} \right\|}_2}}}{{{{\left\| {\boldsymbol{X}} \right\|}_2}}}\;,$
其中$\tilde {\boldsymbol{X}}$是重建图像,${\boldsymbol{X}}$是原始图像,${\left\| \cdot \right\|_2}$表示欧几里得范数。
为了保证实验结果评价的客观性,实验中TV[11],TVa-TVi[18],ETV[20]均采用与文献[20]相同的参数,并采用与文献[39]相同的伪径向采样策略,RTV参数通过分层优化方法与定量指标优化相结合的方式确定。首先,基于文献[19−20]所设定参数的初始范围,借助Shepp-Logan标准图像的网格搜索方法确定参数的取值。然后,固定$\sigma $,$ \mu $取值,图1(a)是不同$\alpha $-$\beta $组合下的RE热力图。最后,固定$\alpha $,$\beta $取值,图1(b)是不同$\sigma $-$ \mu $组合下的$RE$等高线图,为后续图像重建实验的微调提供可靠基准。
另外,噪声环境的图像重建实验主要通过向图像的傅里叶变换域添加随机高斯噪声的方式来实现。基于Matlab的噪声添加方式如式(51)所示。
$ \begin{split}{\boldsymbol{F}} =\;& {\boldsymbol{F}} + \frac{1}{{\sqrt 2 }}\big(std \times randn(size({\boldsymbol{F}})) + \\& std \times 1{\text{i}} \times randn(size({\boldsymbol{F}}))\big)\;,\end{split} $
其中${\boldsymbol{F}}$表示图像${\boldsymbol{X}}$的傅里叶变换,噪声的强度通过标准差$std$进行调整,$randn(\;) $表示生成服从标准正态分布的随机数,1i表示复数虚数。
实验中采用与文献[19-20]相同的样本图像。如图2所示,分别是标准的Shepp-Logan图像、3张自然图像(Cameraman,Clock,Peppers)和2张医学图像(Brain,Spine)。3种图像的大小均为$256 \times 256$像素。1)Shepp-Logan图像是一种合成图像,模拟了人类的头部截面,是医学成像领域中常用的一个标准测试图像;2)自然图像通常包含复杂的纹理和细节,比Shepp-Logan这种合成图像能更好地模拟真实应用场景中的图像数据,本文实验中使用的自然图像来自USC-SIPI数据集[43];3)所选取的Brain图像[44]和Spine图像[45]均为医学磁共振图像,更接近于真实的应用场景。图2(a)~(d)用于比较有效性和高效性;图2(e)(f)用于实用性验证。
本节根据4.1节的4个经典指标,首先对图2所示的经典的Shepp-Logan图像在无噪声干扰情况下对RTV方法的有效性进行评价。RTV方法的参数如表1所示。
图3是应用本文提出的RTV方法对Shepp-Logan图像重建的实验结果。图3(a)表示原始图像,图3(b)表示无噪声条件下的采样掩模,图3(c)表示对图3(a)应用RTV方法的重建图像。从图3所示的重建图像可以看出,在无噪声干扰的条件下,RTV方法在低采样率(3.03%)下重建的图像在视觉上非常接近原始图像,从表2中SSIM,RE,PSNR,RMSE这些评价指标上也可以看出,RTV方法的重建结果具有较高的保真度,这证明了RTV方法在稀疏采样场景中的有效性。
图4是应用所提出的RTV方法分别对Cameraman,Clock,Peppers图像的重建实验结果,RTV方法的参数如表3所示,相应的实验评价指标如表4所示。图4(a)(d)(g)表示原始图像,图4(b)(e)(h)表示不同采样率下的伪径向采样掩膜,图4(c)(f)(i)表示使用所提出的RTV方法重建的图像。从图4所示的重建结果和表4所示的评价指标可以看出,RTV方法对自然图像同样进行了有效重建。但是Peppers图像的重建效果一般,这是因为与Shepp-Logan图像相比,自然图像具有更为复杂的图像结构,同时Peppers图像比Cameraman,Clock图像细节更加丰富,所以较低采样率下重建图像的保真度不如Shepp-Logan图像。
为进一步验证RTV方法的有效性,图5展示了RTV方法分别对Cameraman,Clock,Peppers图像的重建实验的评价指标。实验中对RTV方法模型参数的设置如表3所示。通过观察图5可以看出,在3个实验样本图像上,RTV方法重建图像的SSIM,PSNR均随着采样线条数的增加而逐步上升,而重建图像的RMSE,RE随着采样线条数的增加而逐步减小。综合而言,这说明RTV方法的重建效果随着采样率的增加而逐步提高。可见,RTV方法在稀疏采样条件下能够有效恢复高质量的图像。
将所提出RTV的方法与TV[11],TVa-TVi[18],ETV[20]进行比较。图6~9展示了各方法在样本Shepp-Logan,Cameraman,Clock,Peppers上的实验结果。从图6~9的图像重建结果可以看出,对于4种标准图像样本,TV和TVa-TVi方法重建的图像中存在较多伪影,而ETV方法虽然有所改进,但在均匀区域的平滑性方面表现不足。相比之下,所提出的RTV方法重建的图像可以有效抑制阶梯效应和伪影的产生,同时降低边缘处的对比度损失,保持图像均匀区域的平滑性。导致这一结果的主要原因在于所提出的RTV模型通过使用非凸变形${l_1}$范数对传统TV模型进行改进,从而更充分地挖掘了图像在变分域中的稀疏特性,显著增强了对图像结构和细节的感知能力。此外,RTV模型中的后向扩散机制能够有效恢复图像的对比度,提升重建图像的视觉感观效果。
图10是TV,TVa-TVi,ETV,RTV方法在标准图像样本Shepp-Logan,Cameraman,Clock,Peppers上的重建图像的评价指标。图10(a)是各参与比较方法的SSIM值,其中RTV方法重建的图像具有更高SSIM,表明其在视觉上更加接近原始图像;从图10(b)可以看出,RTV方法重建的图像具有更高的PSNR值,这是因为RTV方法结合了后向扩散处理机制,能够在均匀区域有效平滑噪声,同时保留边缘和纹理信息,所以RTV方法在噪声抑制方面效果更好,使得重建图像的质量更接近于原始图像;图10(c)和图10(d)分别是4种方法重建图像的RMSE和RE,可以看出,RTV方法重建图像的RMSE值和RE最小,这是由于引入的后向扩散处理机制能够增强边缘结构,同时有效降低噪声干扰,减小重建误差。
从图6~10的实验结果综合来看,所提出的RTV方法在视觉和评价指标上的表现优于其他参与比较的方法,这说明RTV方法具有一定的高效性。图11展示了在不同噪声强度下,TV[11],TVa-TVi[18],ETV[20],RTV方法对Shepp-Logan图像重建实验的评价指标对比结果。实验设置伪径向线采样的采样率为6.44%,随机高斯噪声的标准差$std \in \{ 0.1,0.12, 0.14, 0.16, 0.18, 0.2\} $。从图11所示评价指标可以看出,RTV方法的SSIM,PSNR保持最高,RMSE和RE最低;ETV方法的重建结果次之,而TV方法和TVa-TVi方法的重建质量较差,这是因为ETV和RTV中所引入的后向扩散处理机制有效抑制了噪声的干扰,所以在噪声存在的时候,ETV和RTV方法能够更准确地恢复图像的结构和细节信息,相比于未引入后向扩散机制的模型(如TV和TVa-TVi),RTV方法的重建结果具有更高的鲁棒性。进一步说明了RTV方法中后向扩散处理机制的有效性。
本节通过对4.2节所示的2张医学图像Brain,Spine进行图像重建实验,以期证明所提出RTV方法的实用性。RTV方法中的模型参数设置如表5所示。
图12、图13分别展示了参与比较的各方法对Brain,Spine图像的重建结果。从图12、图13可以看到,TV方法和ETV方法重建的图像在边缘处的细节和纹理不够清晰;TVa-TVi方法重建的图像存在较多的伪影;而RTV方法在处理医学图像时能有效抑制重建图像的阶梯效应,降低图像边缘附近的对比度损失。这是由于RTV方法对TV进行了非凸改造并引入后向扩散处理机制,增强了对噪声的抑制能力,同时更贴近${l_1}$范数的稀疏性建模,能够更有效地去除噪声引起的小幅伪影。另外,通过观察图12和图13发现,RTV方法改善了图像中可能出现的伪影问题,并保持了图像均匀区域的平滑性,这是因为后向扩散机制在均匀区域对梯度进行了选择性调节,使图像呈现更自然的过渡效果,而不是简单的强制平滑。这一结果说明RTV方法所重建的图像在视觉上更加接近原始图像,同时从一个侧面展示了RTV方法的实用潜力。
为进一步说明RTV方法的重建能力和效果。图14、图15展示了不同采样率下TV[11],TVa-TVi[18],ETV[20]、RTV方法在实际样本图像Brain,Spine上的各种重建性能指标。从图14、图15可以不难看出,RTV方法在SSIM,PSNR,RMSE,RE这4个评价指标上的表现均优于其他参与比较的方法。结合图12、图13的实验结果进行综合分析,这是因为RTV方法使用变形${l_1}$函数对TV进行了非凸改进,所以RTV利用这一特性能够充分利用图像在变分域的稀疏性,而且引入的后向扩散处理机制有效抑制了重建图像在边缘附近的对比度损失,导致所重建的图像具有更佳的细节保留能力和抗噪声干扰能力。这一实验结果,证实了所提出的RTV方法具有一定的实用性。
本文针对稀疏采样的图像重建应用需求,研究并提出了一种图像重建模型。本模型旨在解决现有基于TV最小化的图像重建模型中存在的阶梯效应和边缘对比度损失问题。通过求解该模型,进一步提出一种RTV图像重建方法。实验结果表明,RTV能充分利用图像在梯度域的稀疏性,有效抑制重建图像的阶梯效应,并降低边缘附近的对比度损失。同时具有较好的抗噪声干扰能力。最后通过实验验证了RTV的有效性、高效性和实用性。然而,RTV尚存在一些不足之处,所提出模型的参数是根据经验取值,如何自适应地确定模型的参数,以及如何将RTV作为深度学习模型的正则化项,探索RTV与深度学习的结合等将是我们下一步的研究工作。
作者贡献声明:张飞翔负责对模型进行数学证明和算法设计,并完成算法的仿真;李志华负责构建模型并对全文进行修改;刘建军负责协助完成模型理论推导和算法代码实现。

参考文献 引证文献
排序方式:
1
Donoho D L. Compressed sensing[J]. IEEE Transactions on Information Theory, 2006, 52(4): 1289−1306
2
陈律君, 肖迪, 余柱阳, 等. 基于秘密共享和压缩感知的通信高效联邦学习[J]. 计算机研究与发展, 2022, 59(11): 2395−2407
Chen Lüjun, Xiao Di, Yu Zhuyang, et al. Communication-efficient federated learning based on secret sharing and compressed sensing[J]. Journal of Computer Research and Development, 2022, 59(11): 2395−2407 (in Chinese)
3
Chambolle A, Pock T. An introduction to continuous optimization for imaging[J]. Acta Numerica, 2016, 25(1): 161−319
4
Rudin L I, Osher S, Fatemi E. Nonlinear total variation based noise removal algorithms[J]. Physica D: Nonlinear Phenomena, 1992, 60(1/2/3/4): 259−268
5
Aharon M, Elad M, Bruckstein A. K-SVD: An algorithm for designing overcomplete dictionaries for sparse representation[J]. IEEE Transactions on Signal Processing, 2006, 54(11): 4311−4322
6
Hoerl A E, Kennard R W. Ridge regression: Biased estimation for nonorthogonal problems[J]. Technometrics, 1970, 12(1): 55−67
7
Candès E, Recht B. Exact matrix completion via convex optimization[J]. Communications of the ACM, 2012, 55(6): 111−119
8
Dong Chao, Loy C C, He Kaiming, et al. Image super-resolution using deep convolutional networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 38(2): 295−307
9
Daubechies I, Defrise M, De Mol C. An iterative thresholding algorithm for linear inverse problems with a sparsity constraint[J]. Communications on Pure and Applied Mathematics, 2004, 57(11): 1413−1457
10
Chambolle A. An algorithm for total variation minimization and applications[J]. Journal of Mathematical Imaging and Vision, 2004, 20(1): 89−97
11
Needell D, Ward R. Stable image reconstruction using total variation minimization[J]. SIAM Journal on Imaging Sciences, 2013, 6(2): 1035−1058
12
Bredies K, Kunisch K, Pock T. Total generalized variation[J]. SIAM Journal on Imaging Sciences, 2010, 3(3): 492−526
13
Zhang Honglu, Tang Liming, Fang Zhuang, et al. Nonconvex and nonsmooth total generalized variation model for image restoration[J]. Signal Processing, 2018, 143(1): 69−85
14
Zhu Wei, Chan T. Image denoising using mean curvature of image surface[J]. SIAM Journal on Imaging Sciences, 2012, 5(1): 1−32
15
Fairag F, Chen Ke, Brito-Loeza C, et al. A two-level method for image denoising and image deblurring models using mean curvature regularization[J]. International Journal of Computer Mathematics, 2022, 99(4): 693−713
16
Deng Liangjian, Glowinski R, Tai Xuecheng. A new operator splitting method for the euler elastica model for image smoothing[J]. SIAM Journal on Imaging Sciences, 2019, 12(2): 1190−1230
17
Zhang Hanming, Wang Linyuan, Duan Yuping, et al. Euler’s elastica strategy for limited-angle computed tomography image reconstruction[J]. IEEE Transactions on Nuclear Science, 2017, 64(8): 2395−2405
18
Lou Yifei, Zeng Tieyong, Osher S, et al. A weighted difference of anisotropic and isotropic total variation model for image processing[J]. SIAM Journal on Imaging Sciences, 2015, 8(3): 1798−1823
19
Huo Limei, Chen Wengu, Ge Huanmin, et al. Stable image reconstruction using transformed total variation minimization[J]. SIAM Journal on Imaging Sciences, 2022, 15(3): 1104−1139
20
An Congpei, Wu Haoning, Yuan Xiaoming. Enhanced total variation minimization for stable image reconstruction[J]. Inverse Problems, 2023, 39(7): 075005
21
Alvarez L, Mazorra L. Signal and image restoration using shock filters and anisotropic diffusion[J]. SIAM Journal on Numerical Analysis, 1994, 31(2): 590−605
22
Osher S, Rudin L I. Feature-oriented image enhancement using shock filters[J]. SIAM Journal on Numerical Analysis, 1990, 27(4): 919−940
23
Welk M, Weickert J, Galić I. Theoretical foundations for spatially discrete 1-D shock filtering[J]. Image and Vision Computing, 2007, 25(4): 455−463
24
Gilboa G, Sochen N, Zeevi Y Y. Forward-and-backward diffusion processes for adaptive image enhancement and denoising[J]. IEEE Transactions on Image Processing, 2002, 11(7): 689−703
25
Welk M, Weickert J, Gilboa G. A discrete theory and efficient algorithms for forward-and-backward diffusion filtering[J]. Journal of Mathematical Imaging and Vision, 2018, 60(9): 1399−1426
26
Welk M, Gilboa G, Weickert J. Theoretical foundations for discrete forward-and-backward diffusion filtering[G]//LNCS 5567: Proc of the 2nd Int Conf on Scale Space and Variational Methods in Computer Vision. Berlin: Springer, 2009: 527−538
27
Tikhonov A N, Arsenin V Y. Solutions of Ill-Posed Problems[M]. Hoboken, NJ: John Wiley &amp; Sons, 1977
28
Mizusawa S, Sei Y, Orihara R, et al. Computed tomography image reconstruction using stacked U-Net[J]. Computerized Medical Imaging and Graphics, 2021, 90(1): 101920
29
Liu Jiaming, Sun Yu, Xu Xiaojian, et al. Image restoration using total variation regularized deep image prior[C]//Proc of the 44th Int Conf on Acoustics, Speech and Signal Processing. Piscataway, NJ: IEEE, 2019: 7715−7719
30
Hussein S A, Tirer T, Giryes R. Image-adaptive GAN based reconstruction[C]//Proc of the 34th Int Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2020: 3121−3129
31
Tao P D, An L T H. Convex analysis approach to DC programming: Theory, algorithms and applications[J]. Acta Mathematica Vietnamica, 1997, 22(1): 289−355
32
Boyd S. Distributed optimization and statistical learning via the alternating direction method of multipliers[J]. Foundations and Trends® in Machine Learning, 2010, 3(1): 1−122
33
Chambolle A, Pock T. Learning consistent discretizations of the total variation[J]. SIAM Journal on Imaging Sciences, 2021, 14(2): 778−813
34
Condat L. Discrete total variation: New definition and minimization[J]. SIAM Journal on Imaging Sciences, 2017, 10(3): 1258−1290
35
Bruckstein A M, Donoho D L, Elad M. From sparse solutions of systems of equations to sparse modeling of signals and images[J]. SIAM Review, 2009, 51(1): 34−81
36
Fan Jianqing, Li Runze. Variable selection via nonconcave penalized likelihood and its oracle properties[J]. Journal of the American Statistical Association, 2001, 96(456): 1348−1360
37
Zhang Shuai, Xin J. Minimization of transformed L_1 penalty: Theory, difference of convex function algorithm, and robust application in compressed sensing[J]. Mathematical Programming, 2018, 169(1): 307−336
38
Zhang Shuai, Xin J. Minimization of transformed l_1 penalty: Closed form representation and iterative thresholding algorithms[J]. Communications in Mathematical Sciences, 2017, 15(2): 511−537
39
Sherry F, Benning M, De los Reyes J C, et al. Learning the sampling pattern for MRI[J]. IEEE Transactions on Medical Imaging, 2020, 39(12): 4310−4321
40
Tao P D, An L T H. A DC optimization algorithm for solving the trust-region subproblem[J]. SIAM Journal on Optimization, 1998, 8(2): 476−505
41
Tibshirani R. Regression shrinkage and selection via the lasso[J]. Journal of the Royal Statistical Society Series B: Statistical Methodology, 1996, 58(1): 267−288
42
Wang Zhou, 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
43
Weber A. The USC-SIPI image database[EB/OL]. University of Southern California, Signal and Image Processing Institute, 1977[2025-05-05]. http://sipi.usc.edu/database/
44
Lalys F, Haegelen C, Ferre J C, et al. Construction and assessment of a 3-T MRI brain template[J]. NeuroImage, 2010, 49(1): 345−354
45
Gonzalez R C, Woods R E. Digital Image Processing[M]. Noida: Pearson Education India, 2009
2026年第63卷第2期
271
129
0
引用本文
BibTeX
文章信息
doi: 10.7544/issn1000-1239.202550014
  • 接收时间:2025-01-10
  • 首发时间:2026-11-06
  • 出版时间:2026-02-01
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2025-01-10
  • 修回日期:2025-07-08
  • 录用日期:2025-07-17
基金
作者信息
    江南大学人工智能与计算机学院 江苏无锡 214122

通讯作者:

李志华()
参考文献
分享链接
https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/10.7544/issn1000-1239.202550014
分享至
全文二维码

扫描看全文

引用本文
BibTeX
本文的引用情况
2种不同金属材料的力学参数
科
Family
属数
Number of
genus
种数
Number of
species
占总种数比例
Percentage of
total species (%)
属
Genus
种数
Number of
species
占总种数比例
Percentage of total
species (%)
鹅膏菌科Amanitaceae 2 11 5.26 鹅膏菌属 Amanita 10 4.78
小菇科 Mycenaceae 2 12 5.74 丝盖伞属 Inocybe 5 2.39
多孔菌科 Polyporaceae 8 14 6.70 蜡蘑属 Laccaria 5 2.39
红菇科 Russulaceae 3 23 11.00 小皮伞属 Marasmius 6 2.87
小菇属 Mycena 11 5.26
光柄菇属 Pluteus 5 2.39
红菇属 Russula 17 8.13
栓菌属 Trametes 5 2.39
关闭全屏