Edge-Guided Shallow and Contextual Deep Feature Learning via Bidirectional Attention for Salient Object Detection in Optical Remote Sensing Images

Kanwal, S; Waheed, N; Rashid, B; Alqahtani, N; Alqahtani, A

Edge-Guided Shallow and Contextual Deep Feature Learning via Bidirectional Attention for Salient Object Detection in Optical Remote Sensing Images

Kanwal, S Waheed, N

Rashid, B Alqahtani, N Alqahtani, A

Permalink

Publisher:: IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
Publication Type:: Journal Article
Citation:: IEEE ACCESS, 2025, 13, (99), pp. 218066-218081
Issue Date:: 2025

Open Access

Copyright Clearance Process

Recently Added
In Progress
Open Access

This item is open access.

Adobe PDF

Download Published versionAdobe PDF (3.8 MB)

View on publisher's site

View statistics

Full metadata record

Field	Value	Language
dc.contributor.author	Kanwal, S
dc.contributor.author	Waheed, N https://orcid.org/0000-0001-5679-5365
dc.contributor.author	Rashid, B
dc.contributor.author	Alqahtani, N
dc.contributor.author	Alqahtani, A
dc.date.accessioned	2026-02-09T05:52:45Z
dc.date.available	2026-02-09T05:52:45Z
dc.date.issued	2025
dc.identifier.citation	IEEE ACCESS, 2025, 13, (99), pp. 218066-218081
dc.identifier.issn	2169-3536
dc.identifier.issn	2169-3536
dc.identifier.uri	http://hdl.handle.net/10453/193116
dc.description.abstract	Salient object detection from optical remote sensing images seeks to detect and segment out objects that stand out visually by mimicking human interpretation. Accurately detecting salient objects in remote sensing imagery is hindered by the broad variability in object appearances and background complexity. Despite recent advancements, there exist two major challenges. Firstly, it remains unclear how to fuse shallow features with deep features optimally. Secondly, defining an optimal strategy for multi-scale feature processing is crucial, as patterns can vary significantly across different levels. We introduce a novel bidirectional attention mechanism to tackle these challenges. Our framework features: (1) a Parallel Convolution-Channel Attention (PCCA) module that boosts edge representation and highlights salient feature channels, and (2) a Holistic Reverse Attention (HRA) module that preserves crucial details in high-level features. Coupled with a multiscale progressive decoder, our method enables precise feature integration. Extensive evaluation on benchmark datasets demonstrates the superiority of our proposed model compared with 17 state-of-the-art (SOTA) models, achieving a significant 11.57% reduction in Mean Absolute Error and setting a new state-of-the-art on S-measure with a 0.4% improvement.
dc.language	English
dc.publisher	IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
dc.relation.ispartof	IEEE ACCESS
dc.relation.isbasedon	10.1109/ACCESS.2025.3648292
dc.rights	info:eu-repo/semantics/openAccess
dc.subject	08 Information and Computing Sciences, 09 Engineering, 10 Technology
dc.subject.classification	40 Engineering
dc.subject.classification	46 Information and computing sciences
dc.title	Edge-Guided Shallow and Contextual Deep Feature Learning via Bidirectional Attention for Salient Object Detection in Optical Remote Sensing Images
dc.type	Journal Article
utslib.citation.volume	13
utslib.for	08 Information and Computing Sciences
utslib.for	09 Engineering
utslib.for	10 Technology
pubs.organisational-group	University of Technology Sydney
pubs.organisational-group	University of Technology Sydney/Faculty of Engineering and Information Technology
utslib.copyright.status	open_access	*
dc.rights.license	This work is licensed under a Creative Commons Attribution 4.0 International License (CC BY 4.0). To view a copy of this license, visit https://creativecommons.org/licenses/by/4.0/
dc.date.updated	2026-02-09T05:52:32Z
pubs.issue	99
pubs.publication-status	Published
pubs.volume	13
utslib.citation.issue	99

Abstract:

Salient object detection from optical remote sensing images seeks to detect and segment out objects that stand out visually by mimicking human interpretation. Accurately detecting salient objects in remote sensing imagery is hindered by the broad variability in object appearances and background complexity. Despite recent advancements, there exist two major challenges. Firstly, it remains unclear how to fuse shallow features with deep features optimally. Secondly, defining an optimal strategy for multi-scale feature processing is crucial, as patterns can vary significantly across different levels. We introduce a novel bidirectional attention mechanism to tackle these challenges. Our framework features: (1) a Parallel Convolution-Channel Attention (PCCA) module that boosts edge representation and highlights salient feature channels, and (2) a Holistic Reverse Attention (HRA) module that preserves crucial details in high-level features. Coupled with a multiscale progressive decoder, our method enables precise feature integration. Extensive evaluation on benchmark datasets demonstrates the superiority of our proposed model compared with 17 state-of-the-art (SOTA) models, achieving a significant 11.57% reduction in Mean Absolute Error and setting a new state-of-the-art on S-measure with a 0.4% improvement.

Please use this identifier to cite or link to this item:

http://hdl.handle.net/10453/193116