ATT-CR: Adaptive Triangular Transformer for Cloud Removal

Cloud removal aims to accurately reconstruct the ground objects obscured by clouds in remote sensing images. Existing transformer-based methods utilizing self-attention have shown impressive results by effectively modeling long-range dependencies in cloudy images. However, they suffer from the follo...

Full description

Saved in:
Bibliographic Details
Main Authors: Yang Wu, Ye Deng, Pengna Li, Wenli Huang, Kangyi Wu, Xiaomeng Xin, Jinjun Wang
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/11119343/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849223120426106880
author Yang Wu
Ye Deng
Pengna Li
Wenli Huang
Kangyi Wu
Xiaomeng Xin
Jinjun Wang
author_facet Yang Wu
Ye Deng
Pengna Li
Wenli Huang
Kangyi Wu
Xiaomeng Xin
Jinjun Wang
author_sort Yang Wu
collection DOAJ
description Cloud removal aims to accurately reconstruct the ground objects obscured by clouds in remote sensing images. Existing transformer-based methods utilizing self-attention have shown impressive results by effectively modeling long-range dependencies in cloudy images. However, they suffer from the following issues: 1) the high computational complexity of self-attention limits scalability; 2) treating both cloudy and clean pixels as valid within the attention computation brings disturbances in subsequent layers, leading to suboptimal performance. To address these challenges, we propose the adaptive triangular transformer for cloud removal (ATT-CR), a model that effectively reduces computational costs and mitigates interference from cloudy pixels. Specifically, it consists of two core components: Triangular attention (TAN) and feature selected gating module (FSGM). TAN employs lower and upper triangular matrices to approximate softmax attention with <inline-formula><tex-math notation="LaTeX">$\mathcal {O}(N)$</tex-math></inline-formula> computational complexity, significantly reducing the computational costs. The FSGM, on the other hand, integrates with TAN to adaptively distinguish between cloudy and clean features, which minimizes the introduction of invalid information into subsequent layers. Extensive experiments on cloud removal benchmarks demonstrate that ATT-CR delivers superior performance compared to existing methods.
format Article
id doaj-art-453f0a2964e1467daeb224efc84e17b4
institution Kabale University
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-453f0a2964e1467daeb224efc84e17b42025-08-25T23:00:17ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing1939-14042151-15352025-01-0118205952061010.1109/JSTARS.2025.359685611119343ATT-CR: Adaptive Triangular Transformer for Cloud RemovalYang Wu0https://orcid.org/0009-0006-3944-6415Ye Deng1https://orcid.org/0000-0003-4616-3318Pengna Li2https://orcid.org/0000-0002-8477-8340Wenli Huang3https://orcid.org/0000-0002-8636-044XKangyi Wu4Xiaomeng Xin5https://orcid.org/0009-0006-6307-1465Jinjun Wang6https://orcid.org/0000-0002-9434-0617Institute of Artificial Intelligence and Robotics, Xi&#x2019;an Jiaotong University, Xi&#x2019;an, ChinaEngineering Research Center of Intelligent Finance, Ministry of Education, School of Computing and Artificial Intelligence, Southwestern University of Finance and Economics, Chengdu, ChinaInstitute of Artificial Intelligence and Robotics, Xi&#x2019;an Jiaotong University, Xi&#x2019;an, ChinaNingbo University of Technology, Ningbo, ChinaInstitute of Artificial Intelligence and Robotics, Xi&#x2019;an Jiaotong University, Xi&#x2019;an, ChinaInstitute of Artificial Intelligence and Robotics, Xi&#x2019;an Jiaotong University, Xi&#x2019;an, ChinaInstitute of Artificial Intelligence and Robotics, Xi&#x2019;an Jiaotong University, Xi&#x2019;an, ChinaCloud removal aims to accurately reconstruct the ground objects obscured by clouds in remote sensing images. Existing transformer-based methods utilizing self-attention have shown impressive results by effectively modeling long-range dependencies in cloudy images. However, they suffer from the following issues: 1) the high computational complexity of self-attention limits scalability; 2) treating both cloudy and clean pixels as valid within the attention computation brings disturbances in subsequent layers, leading to suboptimal performance. To address these challenges, we propose the adaptive triangular transformer for cloud removal (ATT-CR), a model that effectively reduces computational costs and mitigates interference from cloudy pixels. Specifically, it consists of two core components: Triangular attention (TAN) and feature selected gating module (FSGM). TAN employs lower and upper triangular matrices to approximate softmax attention with <inline-formula><tex-math notation="LaTeX">$\mathcal {O}(N)$</tex-math></inline-formula> computational complexity, significantly reducing the computational costs. The FSGM, on the other hand, integrates with TAN to adaptively distinguish between cloudy and clean features, which minimizes the introduction of invalid information into subsequent layers. Extensive experiments on cloud removal benchmarks demonstrate that ATT-CR delivers superior performance compared to existing methods.https://ieeexplore.ieee.org/document/11119343/Adaptive feature selectioncloud removalimage reconstructionremote sensing imagestriangular attention (TAN)
spellingShingle Yang Wu
Ye Deng
Pengna Li
Wenli Huang
Kangyi Wu
Xiaomeng Xin
Jinjun Wang
ATT-CR: Adaptive Triangular Transformer for Cloud Removal
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Adaptive feature selection
cloud removal
image reconstruction
remote sensing images
triangular attention (TAN)
title ATT-CR: Adaptive Triangular Transformer for Cloud Removal
title_full ATT-CR: Adaptive Triangular Transformer for Cloud Removal
title_fullStr ATT-CR: Adaptive Triangular Transformer for Cloud Removal
title_full_unstemmed ATT-CR: Adaptive Triangular Transformer for Cloud Removal
title_short ATT-CR: Adaptive Triangular Transformer for Cloud Removal
title_sort att cr adaptive triangular transformer for cloud removal
topic Adaptive feature selection
cloud removal
image reconstruction
remote sensing images
triangular attention (TAN)
url https://ieeexplore.ieee.org/document/11119343/
work_keys_str_mv AT yangwu attcradaptivetriangulartransformerforcloudremoval
AT yedeng attcradaptivetriangulartransformerforcloudremoval
AT pengnali attcradaptivetriangulartransformerforcloudremoval
AT wenlihuang attcradaptivetriangulartransformerforcloudremoval
AT kangyiwu attcradaptivetriangulartransformerforcloudremoval
AT xiaomengxin attcradaptivetriangulartransformerforcloudremoval
AT jinjunwang attcradaptivetriangulartransformerforcloudremoval