DRHDR: A Dual branch Residual Network for Multi-Bracket High Dynamic Range Imaging

Juan Marin-Vega, Michael Sloth, Peter Schneider-Kamp, Richard Rottger

Research output: Contribution to journalConference articleResearchpeer-review

Abstract

We introduce DRHDR, a Dual branch Residual Convolutional Neural Network for Multi-Bracket HDR Imaging. To address the challenges of fusing multiple brackets from dynamic scenes, we propose an efficient dual branch network that operates on two different resolutions. The full resolution branch uses a Deformable Convolutional Block to align features and retain high-frequency details. A low resolution branch with a Spatial Attention Block aims to attend wanted areas from the non-reference brackets, and suppress displaced features that could incur on ghosting artifacts. By using a dual branch approach we are able to achieve high quality results while constraining the computational resources required to estimate the HDR results.

Original languageEnglish
JournalIEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops
Pages (from-to)843-851
Number of pages9
ISSN2160-7508
DOIs
Publication statusPublished - 2022
Event2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2022 - New Orleans, United States
Duration: 19. Jun 202220. Jun 2022

Conference

Conference2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2022
Country/TerritoryUnited States
CityNew Orleans
Period19/06/202220/06/2022

Bibliographical note

Funding Information:
This work was supported by the Industrial Ph.D. program’s financial support from Innovation Fund Denmark, through the project AIERE (Contract-No: 9065-00099B).

Fingerprint

Dive into the research topics of 'DRHDR: A Dual branch Residual Network for Multi-Bracket High Dynamic Range Imaging'. Together they form a unique fingerprint.

Cite this