Reference-Guided Texture Transfer with Deformable Convolutions for Indoor Image Dehazing

Reference-Guided Texture Transfer with Deformable Convolutions for Indoor Image Dehazing | 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 Article Reference-Guided Texture Transfer with Deformable Convolutions for Indoor Image Dehazing Esteban Reyes-Saldaña, Mariano Rivera This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8564433/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 10 You are reading this latest preprint version Abstract Purpose : Image dehazing is a vital image restoration task that aims to recover visual clarity and contrast from images degraded by haze. While much of the existing literature focuses on outdoor scenes such as landscapes or entertainment photography, indoor haze poses a distinct and critical challenge, often caused by smoke in enclosed environments, such as during fire emergencies. In these scenarios, faithful restoration is paramount: the reconstructed image must remain as close as possible to the ground truth, without introducing artificial textures or unrealistic elements, since accuracy can directly impact decision-making. Methods : In this paper, we propose DTTN, an enhanced transformer-based framework that advances our earlier Texture Transfer Dehazing Network (TTDN) by incorporating deformable convolutions and a streamlined architecture designed to better capture nonlocal dependencies and spatially adaptive features. DTTN retains the core innovation of the Reference Super-Resolution (RefSR) paradigm by leveraging a high-quality reference image to guide the reconstruction of fine textures in hazy images. Our pipeline extracts deep features from both the input hazy image and a clean high-resolution reference using a modified VGG19 backbone, transforming them into patch-based representations. Using Deformable Convolutional Networks (DCNs), the model dynamically aligns relevant textures from the reference, enhancing spatial correspondence. Furthermore, we introduce a Gradient Density Enhancement Module, which leverages edge and structural cues to further improve restoration fidelity. Results : We evaluated DTTN on the RESIDE-Indoor dataset, reporting new benchmark results. Quantitative evaluations demonstrate that DTTN achieves or exceeds state-of-the-art performance on standard metrics such as PSNR and SSIM, while also improving computational efficiency via optimized patch sizes and strides in the texture transfer module. Qualitative comparisons highlight the ability of DTTN to preserve fine textures and structural consistency across a wide range of indoor haze scenarios. Conclusion : Overall, our findings highlight DTTN as an effective and efficient solution for faithful indoor image dehazing, with strong potential for deployment in safety-critical vision applications. Physical sciences/Engineering Physical sciences/Mathematics and computing Image Dehazing Texture Transformers Texture Transference Deformable Convolutions Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 20 Mar, 2026 Reviews received at journal 16 Mar, 2026 Reviewers agreed at journal 05 Mar, 2026 Reviews received at journal 03 Mar, 2026 Reviewers agreed at journal 03 Mar, 2026 Reviewers invited by journal 03 Mar, 2026 Editor invited by journal 19 Jan, 2026 Editor assigned by journal 16 Jan, 2026 Submission checks completed at journal 14 Jan, 2026 First submitted to journal 14 Jan, 2026 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-8564433","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":601019063,"identity":"88cec02d-41bd-4d99-95e8-4007933c5392","order_by":0,"name":"Esteban Reyes-Saldaña","email":"","orcid":"","institution":"Centro de Investigacion en Matematicas AC","correspondingAuthor":false,"prefix":"","firstName":"Esteban","middleName":"","lastName":"Reyes-Saldaña","suffix":""},{"id":601019065,"identity":"3b97cbbd-a8f4-49f2-8c7f-f699745cf495","order_by":1,"name":"Mariano Rivera","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAArklEQVRIiWNgGAWjYFACxgYQKcfATKoWY7CWA6TYlQjWSJQW3dmH2yQ+7rFL72/nTvz8gWEbRC8+YHYusU1yxrPk3BmHeTdLHGC4bUzQFrMzjM3GPAeYczcw824AaZEjTsufA/XpBsy8m38AtfAQo6XxMcOBwwlALduItqXxYc+B44ZAv2yzOGNAlF/YHxz4caBanr//7OYbFRW3CYcYGjAgUf0oGAWjYBSMAuwAABCyPo9pkiExAAAAAElFTkSuQmCC","orcid":"","institution":"Centro de Investigacion en Matematicas AC","correspondingAuthor":true,"prefix":"","firstName":"Mariano","middleName":"","lastName":"Rivera","suffix":""}],"badges":[],"createdAt":"2026-01-09 22:08:13","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8564433/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8564433/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":104402194,"identity":"6293b05e-4dc8-4104-99d7-fd6988288890","added_by":"auto","created_at":"2026-03-11 12:14:37","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2060315,"visible":true,"origin":"","legend":"","description":"","filename":"DehazeCIS3.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8564433/v1_covered_a1d590e8-7265-466b-b36e-62eda56b1486.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Reference-Guided Texture Transfer with Deformable Convolutions for Indoor Image Dehazing","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Image Dehazing, Texture Transformers, Texture Transference, Deformable Convolutions","lastPublishedDoi":"10.21203/rs.3.rs-8564433/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8564433/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003ePurpose\u003c/strong\u003e: Image dehazing is a vital image restoration task that aims to recover visual clarity and contrast from images degraded by haze. While much of the existing literature focuses on outdoor scenes such as landscapes or entertainment photography, indoor haze poses a distinct and critical challenge, often caused by smoke in enclosed environments, such as during fire emergencies. In these scenarios, faithful restoration is paramount: the reconstructed image must remain as close as possible to the ground truth, without introducing artificial textures or unrealistic elements, since accuracy can directly impact decision-making.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods\u003c/strong\u003e: In this paper, we propose DTTN, an enhanced transformer-based framework that advances our earlier Texture Transfer Dehazing Network (TTDN) by incorporating deformable convolutions and a streamlined architecture designed to better capture nonlocal dependencies and spatially adaptive features. DTTN retains the core innovation of the Reference Super-Resolution (RefSR) paradigm by leveraging a high-quality reference image to guide the reconstruction of fine textures in hazy images. Our pipeline extracts deep features from both the input hazy image and a clean high-resolution reference using a modified VGG19 backbone, transforming them into patch-based representations. Using Deformable Convolutional Networks (DCNs), the model dynamically aligns relevant textures from the reference, enhancing spatial correspondence. Furthermore, we introduce a Gradient Density Enhancement Module, which leverages edge and structural cues to further improve restoration fidelity.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults\u003c/strong\u003e: We evaluated DTTN on the RESIDE-Indoor dataset, reporting new benchmark results. Quantitative evaluations demonstrate that DTTN achieves or exceeds state-of-the-art performance on standard metrics such as PSNR and SSIM, while also improving computational efficiency via optimized patch sizes and strides in the texture transfer module. Qualitative comparisons highlight the ability of DTTN to preserve fine textures and structural consistency across a wide range of indoor haze scenarios.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion\u003c/strong\u003e: Overall, our findings highlight DTTN as an effective and efficient solution for faithful indoor image dehazing, with strong potential for deployment in safety-critical vision applications.\u003c/p\u003e","manuscriptTitle":"Reference-Guided Texture Transfer with Deformable Convolutions for Indoor Image Dehazing","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-06 04:29:09","doi":"10.21203/rs.3.rs-8564433/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-03-20T13:41:57+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-03-16T15:58:15+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"13586749349000698346164060740296555393","date":"2026-03-05T06:00:42+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-03-03T13:18:56+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"259020183474858214916942877568930282798","date":"2026-03-03T10:11:36+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-03-03T05:17:40+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-01-19T15:17:52+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-01-16T09:25:27+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-01-15T02:58:38+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2026-01-15T02:56:59+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"0c7bac02-b855-4910-b438-56fbbc82fede","owner":[],"postedDate":"March 6th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":63961238,"name":"Physical sciences/Engineering"},{"id":63961239,"name":"Physical sciences/Mathematics and computing"}],"tags":[],"updatedAt":"2026-05-19T03:54:31+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-06 04:29:09","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8564433","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8564433","identity":"rs-8564433","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}