Dynamic Bidirectional Feature Enhancement Network for Thin Cloud Removal in Remote Sensing Images

Existing thin cloud removal methods primarily rely on generative paradigms or discriminative paradigms. Generative paradigms often suffer from training instability, while discriminative paradigms exhibit insufficient feature representation, and their loss strategies lack physical consistency, result...

Full description

Saved in:
Bibliographic Details
Main Authors: Yu Wang, Hao Chen, Ye Zhang, Guozheng Li
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10994332/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849761641580724224
author Yu Wang
Hao Chen
Ye Zhang
Guozheng Li
author_facet Yu Wang
Hao Chen
Ye Zhang
Guozheng Li
author_sort Yu Wang
collection DOAJ
description Existing thin cloud removal methods primarily rely on generative paradigms or discriminative paradigms. Generative paradigms often suffer from training instability, while discriminative paradigms exhibit insufficient feature representation, and their loss strategies lack physical consistency, resulting in suboptimal performance. To address these issues, we propose a dynamic bidirectional feature enhancement network for thin cloud removal in optical remote sensing images. First, we design a multidimensional attention module comprising skip-recursive dilated convolution-based spatial attention module and frequency-domain-driven channel attention module. These modules efficiently capture long-range contextual dependencies and suppress cloud noise interference. Next, an adaptive local feature enhancement block is constructed using cross-fusion and adjacent feature propagation between dynamic convolutions, aimed at enhancing the ability of model to recover details. Then, we present a dynamic enhancement-based bidirectional information flow module to model the dynamic interaction between multitask features, guiding detail recovery and feedback for optimized cloud removal features. Finally, we design a physics-aware joint loss function incorporating atmospheric light consistency constraints to ensure the physical authenticity of cloud-free images. Evaluation on the RICE and WHUS2-CR datasets demonstrates that the proposed method is superior to compared methods in thin cloud removal and can improve the performance of baseline.
format Article
id doaj-art-e3cadeaffdae43809dce9f6fcc38153a
institution DOAJ
issn 1939-1404
2151-1535
language English
publishDate 2025-01-01
publisher IEEE
record_format Article
series IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
spelling doaj-art-e3cadeaffdae43809dce9f6fcc38153a2025-08-20T03:05:57ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing1939-14042151-15352025-01-0118132811329510.1109/JSTARS.2025.356843710994332Dynamic Bidirectional Feature Enhancement Network for Thin Cloud Removal in Remote Sensing ImagesYu Wang0https://orcid.org/0009-0004-4530-4152Hao Chen1https://orcid.org/0000-0002-1837-3986Ye Zhang2https://orcid.org/0000-0001-8721-4535Guozheng Li3Department of Information Engineering, Harbin Institute of Technology, Harbin, ChinaDepartment of Information Engineering, Harbin Institute of Technology, Harbin, ChinaDepartment of Information Engineering, Harbin Institute of Technology, Harbin, ChinaDepartment of Information Engineering, Harbin Institute of Technology, Harbin, ChinaExisting thin cloud removal methods primarily rely on generative paradigms or discriminative paradigms. Generative paradigms often suffer from training instability, while discriminative paradigms exhibit insufficient feature representation, and their loss strategies lack physical consistency, resulting in suboptimal performance. To address these issues, we propose a dynamic bidirectional feature enhancement network for thin cloud removal in optical remote sensing images. First, we design a multidimensional attention module comprising skip-recursive dilated convolution-based spatial attention module and frequency-domain-driven channel attention module. These modules efficiently capture long-range contextual dependencies and suppress cloud noise interference. Next, an adaptive local feature enhancement block is constructed using cross-fusion and adjacent feature propagation between dynamic convolutions, aimed at enhancing the ability of model to recover details. Then, we present a dynamic enhancement-based bidirectional information flow module to model the dynamic interaction between multitask features, guiding detail recovery and feedback for optimized cloud removal features. Finally, we design a physics-aware joint loss function incorporating atmospheric light consistency constraints to ensure the physical authenticity of cloud-free images. Evaluation on the RICE and WHUS2-CR datasets demonstrates that the proposed method is superior to compared methods in thin cloud removal and can improve the performance of baseline.https://ieeexplore.ieee.org/document/10994332/Adaptive local feature enhancementbidirectional information flowmultidimensional attentionphysics-aware joint
spellingShingle Yu Wang
Hao Chen
Ye Zhang
Guozheng Li
Dynamic Bidirectional Feature Enhancement Network for Thin Cloud Removal in Remote Sensing Images
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Adaptive local feature enhancement
bidirectional information flow
multidimensional attention
physics-aware joint
title Dynamic Bidirectional Feature Enhancement Network for Thin Cloud Removal in Remote Sensing Images
title_full Dynamic Bidirectional Feature Enhancement Network for Thin Cloud Removal in Remote Sensing Images
title_fullStr Dynamic Bidirectional Feature Enhancement Network for Thin Cloud Removal in Remote Sensing Images
title_full_unstemmed Dynamic Bidirectional Feature Enhancement Network for Thin Cloud Removal in Remote Sensing Images
title_short Dynamic Bidirectional Feature Enhancement Network for Thin Cloud Removal in Remote Sensing Images
title_sort dynamic bidirectional feature enhancement network for thin cloud removal in remote sensing images
topic Adaptive local feature enhancement
bidirectional information flow
multidimensional attention
physics-aware joint
url https://ieeexplore.ieee.org/document/10994332/
work_keys_str_mv AT yuwang dynamicbidirectionalfeatureenhancementnetworkforthincloudremovalinremotesensingimages
AT haochen dynamicbidirectionalfeatureenhancementnetworkforthincloudremovalinremotesensingimages
AT yezhang dynamicbidirectionalfeatureenhancementnetworkforthincloudremovalinremotesensingimages
AT guozhengli dynamicbidirectionalfeatureenhancementnetworkforthincloudremovalinremotesensingimages