Enhancing Image Restoration Performance with Hierarchical Swin Transformers

Enhancing Image Restoration Performance with Hierarchical Swin Transformers | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Enhancing Image Restoration Performance with Hierarchical Swin Transformers Rahul Dhuture This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7009485/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Image restoration is a critical task in computer vision, addressing challenges such as image denoising, super-resolution, and compression artifact removal. While convolutional neural networks (CNNs) have traditionally been the backbone of image restoration techniques, they often struggle to model long-range dependencies due to their inherently local receptive fields. In recent years, transformer-based architectures have gained attention for their ability to capture global context effectively. This paper presents a comprehensive study of SwinIR, an advanced image restoration framework based on the Swin Transformer. SwinIR introduces shifted window mechanisms and a hierarchical architecture, allowing efficient modeling of both local and long-range image features. We explore how these innovations contribute to improved restoration performance and better generalization across diverse image degradation scenarios. Extensive experiments are conducted on standard benchmark datasets, covering a wide range of restoration tasks. Our results show that SwinIR consistently surpasses state-of-the-art CNN-based methods in terms of both quantitative metrics, such as PSNR and SSIM, and qualitative visual quality. The findings suggest that Swin Transformer-based models offer a powerful alternative to traditional approaches, paving the way for more accurate and scalable image restoration solutions. Artificial Intelligence and Machine Learning Deep learning Image processing Image restoration Computer vision Transformer architecture SwinIR Full Text Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7009485","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":478364123,"identity":"80223a23-6a3c-4570-a7a4-5cbfe1c9211b","order_by":0,"name":"Rahul Dhuture","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAwklEQVRIiWNgGAWjYHACMwjF3gAkDCyI1ZIApHgOgLRIkKJFAkQwEKHFfNrhbY95fxzO55d8fnXDjwIJBv727gS8WmRup5Ub8yQctpw5O6fsZg/QYRJnzm7Aq0VCOsdMGqjFwOB2TtoNHqAWA4lcIrXY3zyTdvMPSVoMJNiP3SbSlrQyyTlp6QYSZ3LYbssYSPAQ4ZfkbRJvbKwN+NuPP7v55o+NHH97L34tSIDHAEwSqxwE2B+QonoUjIJRMApGEAAAxQpAaEeKYfwAAAAASUVORK5CYII=","orcid":"","institution":"Ramdeobaba University Nagpur","correspondingAuthor":true,"prefix":"","firstName":"Rahul","middleName":"","lastName":"Dhuture","suffix":""}],"badges":[],"createdAt":"2025-06-30 10:31:55","currentVersionCode":1,"declarations":{"humanSubjects":true,"vertebrateSubjects":true,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":true,"humanSubjectConsent":true,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":true},"doi":"10.21203/rs.3.rs-7009485/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7009485/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":85852225,"identity":"cf0228d9-a05b-4640-884f-5fece93ecfa8","added_by":"auto","created_at":"2025-07-02 10:30:41","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":407025,"visible":true,"origin":"","legend":"","description":"","filename":"SpringerMachineLearning.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7009485/v1_covered_98c3115f-c5c0-4f7e-a0f4-b4820bf79d5f.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003e\u003cstrong\u003eEnhancing Image Restoration Performance with Hierarchical Swin Transformers\u003c/strong\u003e\u003c/p\u003e","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Ramdeobaba University","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Deep learning, Image processing, Image restoration, Computer vision, Transformer architecture, SwinIR","lastPublishedDoi":"10.21203/rs.3.rs-7009485/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7009485/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eImage restoration is a critical task in computer vision, addressing challenges such as image denoising, super-resolution, and compression artifact removal. While convolutional neural networks (CNNs) have traditionally been the backbone of image restoration techniques, they often struggle to model long-range dependencies due to their inherently local receptive fields. In recent years, transformer-based architectures have gained attention for their ability to capture global context effectively. This paper presents a comprehensive study of SwinIR, an advanced image restoration framework based on the Swin Transformer. SwinIR introduces shifted window mechanisms and a hierarchical architecture, allowing efficient modeling of both local and long-range image features. We explore how these innovations contribute to improved restoration performance and better generalization across diverse image degradation scenarios. Extensive experiments are conducted on standard benchmark datasets, covering a wide range of restoration tasks. Our results show that SwinIR consistently surpasses state-of-the-art CNN-based methods in terms of both quantitative metrics, such as PSNR and SSIM, and qualitative visual quality. The findings suggest that Swin Transformer-based models offer a powerful alternative to traditional approaches, paving the way for more accurate and scalable image restoration solutions.\u003c/p\u003e","manuscriptTitle":"Enhancing Image Restoration Performance with Hierarchical Swin Transformers","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-07-02 09:58:36","doi":"10.21203/rs.3.rs-7009485/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"862bcbd9-8ad9-4ee8-b61e-3adfd4917454","owner":[],"postedDate":"July 2nd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":50787113,"name":"Artificial Intelligence and Machine Learning"}],"tags":[],"updatedAt":"2025-07-02T09:58:36+00:00","versionOfRecord":[],"versionCreatedAt":"2025-07-02 09:58:36","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7009485","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7009485","identity":"rs-7009485","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}