Cross-frame detail compensation network for ghost-free high dynamic range imaging.
Saved in:
| Title: | Cross-frame detail compensation network for ghost-free high dynamic range imaging. |
|---|---|
| Authors: | Li, Qiang1 (AUTHOR) liqiang@tju.edu.cn, Sun, Yu1 (AUTHOR) 2023232106@tju.edu.cn, Pan, Zhou2 (AUTHOR) panli_an@163.com, Xu, Zibo1 (AUTHOR) xzb6666@tju.edu.cn, Wen, Xin3 (AUTHOR) wenxin113@163.com |
| Source: | Multimedia Systems. Apr2026, Vol. 32 Issue 2, p1-14. 14p. |
| Subjects: | High dynamic range imaging, Image reconstruction, Image processing, Deep learning |
| Abstract: | High dynamic range (HDR) imaging typically merges several low dynamic range (LDR) images to generate a ghost-free HDR image with rich details in both bright and dark regions. Previous methods primarily focused on aligning LDR frames and deghosting to produce ghost-free HDR images. However, they fail to effectively reconstruct fine-grained details from multiple LDR frames and tend to produce ghosting artifacts or distortion in regions with significant motion or over-exposure. To this end, we propose a Cross-Frame Detail Compensation Network for Ghost-free HDR imaging, called CDCNet, to reconstruct high-quality HDR images with rich multi-frame details while significantly reducing ghosting artifacts. Our CDCNet consists of a detail compensation network (DCN) for integrating fine-grained multi-frame details and a feature refinement network (FRN) for deghosting. Specifically, the DCN utilizes a dual-branch alignment module, comprising a Detail Compensation module and a Global Alignment module, to align LDR features and capture complementary cross-frame details. Next, the FRN includes a spatially-refined deghosting (SRD) block with a dual-branch architecture for spatially-enhanced deghosting. The SRD block suppresses ghosting artifacts exploiting a deformable re-attention Transformer encoder, while refining spatial details with a spatial feature extractor. We evaluate our method on three benchmark HDR datasets. Extensive experiments validate the effectiveness of our method in generating realistic and detailed HDR images, both quantitatively and qualitatively, outperforming existing methods. [ABSTRACT FROM AUTHOR] |
| Copyright of Multimedia Systems is the property of Springer Nature and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Database: | Engineering Source |
| FullText | Text: Availability: 0 |
|---|---|
| Header | DbId: egs DbLabel: Engineering Source An: 191288862 AccessLevel: 6 PubType: Academic Journal PubTypeId: academicJournal PreciseRelevancyScore: 0 |
| IllustrationInfo | |
| Items | – Name: Title Label: Title Group: Ti Data: Cross-frame detail compensation network for ghost-free high dynamic range imaging. – Name: Author Label: Authors Group: Au Data: <searchLink fieldCode="AR" term="%22Li%2C+Qiang%22">Li, Qiang</searchLink><relatesTo>1</relatesTo> (AUTHOR)<i> liqiang@tju.edu.cn</i><br /><searchLink fieldCode="AR" term="%22Sun%2C+Yu%22">Sun, Yu</searchLink><relatesTo>1</relatesTo> (AUTHOR)<i> 2023232106@tju.edu.cn</i><br /><searchLink fieldCode="AR" term="%22Pan%2C+Zhou%22">Pan, Zhou</searchLink><relatesTo>2</relatesTo> (AUTHOR)<i> panli_an@163.com</i><br /><searchLink fieldCode="AR" term="%22Xu%2C+Zibo%22">Xu, Zibo</searchLink><relatesTo>1</relatesTo> (AUTHOR)<i> xzb6666@tju.edu.cn</i><br /><searchLink fieldCode="AR" term="%22Wen%2C+Xin%22">Wen, Xin</searchLink><relatesTo>3</relatesTo> (AUTHOR)<i> wenxin113@163.com</i> – Name: TitleSource Label: Source Group: Src Data: <searchLink fieldCode="JN" term="%22Multimedia+Systems%22">Multimedia Systems</searchLink>. Apr2026, Vol. 32 Issue 2, p1-14. 14p. – Name: Subject Label: Subjects Group: Su Data: <searchLink fieldCode="DE" term="%22High+dynamic+range+imaging%22">High dynamic range imaging</searchLink><br /><searchLink fieldCode="DE" term="%22Image+reconstruction%22">Image reconstruction</searchLink><br /><searchLink fieldCode="DE" term="%22Image+processing%22">Image processing</searchLink><br /><searchLink fieldCode="DE" term="%22Deep+learning%22">Deep learning</searchLink> – Name: Abstract Label: Abstract Group: Ab Data: High dynamic range (HDR) imaging typically merges several low dynamic range (LDR) images to generate a ghost-free HDR image with rich details in both bright and dark regions. Previous methods primarily focused on aligning LDR frames and deghosting to produce ghost-free HDR images. However, they fail to effectively reconstruct fine-grained details from multiple LDR frames and tend to produce ghosting artifacts or distortion in regions with significant motion or over-exposure. To this end, we propose a Cross-Frame Detail Compensation Network for Ghost-free HDR imaging, called CDCNet, to reconstruct high-quality HDR images with rich multi-frame details while significantly reducing ghosting artifacts. Our CDCNet consists of a detail compensation network (DCN) for integrating fine-grained multi-frame details and a feature refinement network (FRN) for deghosting. Specifically, the DCN utilizes a dual-branch alignment module, comprising a Detail Compensation module and a Global Alignment module, to align LDR features and capture complementary cross-frame details. Next, the FRN includes a spatially-refined deghosting (SRD) block with a dual-branch architecture for spatially-enhanced deghosting. The SRD block suppresses ghosting artifacts exploiting a deformable re-attention Transformer encoder, while refining spatial details with a spatial feature extractor. We evaluate our method on three benchmark HDR datasets. Extensive experiments validate the effectiveness of our method in generating realistic and detailed HDR images, both quantitatively and qualitatively, outperforming existing methods. [ABSTRACT FROM AUTHOR] – Name: AbstractSuppliedCopyright Label: Group: Ab Data: <i>Copyright of Multimedia Systems is the property of Springer Nature and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract.</i> (Copyright applies to all Abstracts.) |
| PLink | https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=egs&AN=191288862 |
| RecordInfo | BibRecord: BibEntity: Identifiers: – Type: doi Value: 10.1007/s00530-025-02160-2 Languages: – Code: eng Text: English PhysicalDescription: Pagination: PageCount: 14 StartPage: 1 Subjects: – SubjectFull: High dynamic range imaging Type: general – SubjectFull: Image reconstruction Type: general – SubjectFull: Image processing Type: general – SubjectFull: Deep learning Type: general Titles: – TitleFull: Cross-frame detail compensation network for ghost-free high dynamic range imaging. Type: main BibRelationships: HasContributorRelationships: – PersonEntity: Name: NameFull: Li, Qiang – PersonEntity: Name: NameFull: Sun, Yu – PersonEntity: Name: NameFull: Pan, Zhou – PersonEntity: Name: NameFull: Xu, Zibo – PersonEntity: Name: NameFull: Wen, Xin IsPartOfRelationships: – BibEntity: Dates: – D: 01 M: 04 Text: Apr2026 Type: published Y: 2026 Identifiers: – Type: issn-print Value: 09424962 Numbering: – Type: volume Value: 32 – Type: issue Value: 2 Titles: – TitleFull: Multimedia Systems Type: main |
| ResultId | 1 |
