An Efficient and Low-Complexity Transformer-Based Deep Learning Framework for High-Dynamic-Range Image Reconstruction

High-dynamic-range (HDR) image reconstruction involves creating an HDR image from multiple low-dynamic-range images as input, providing a computational solution to enhance image quality. This task presents several challenges, such as frame misalignment, overexposure, and motion, which are addressed...

Full description

Saved in:
Bibliographic Details
Main Authors: Josue Lopez-Cabrejos, Thuanne Paixão, Ana Beatriz Alvarez, Diodomiro Baldomero Luque
Format: Article
Language:English
Published: MDPI AG 2025-02-01
Series:Sensors
Subjects:
Online Access:https://www.mdpi.com/1424-8220/25/5/1497
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850053040427499520
author Josue Lopez-Cabrejos
Thuanne Paixão
Ana Beatriz Alvarez
Diodomiro Baldomero Luque
author_facet Josue Lopez-Cabrejos
Thuanne Paixão
Ana Beatriz Alvarez
Diodomiro Baldomero Luque
author_sort Josue Lopez-Cabrejos
collection DOAJ
description High-dynamic-range (HDR) image reconstruction involves creating an HDR image from multiple low-dynamic-range images as input, providing a computational solution to enhance image quality. This task presents several challenges, such as frame misalignment, overexposure, and motion, which are addressed using deep learning algorithms. In this context, various architectures with different approaches exist, such as convolutional neural networks, diffusion networks, generative adversarial networks, and Transformer-based architectures, with the latter offering the best quality but at a high computational cost. This paper proposes an HDR reconstruction architecture using a Transformer-based approach to achieve results competitive with the state of the art while reducing computational cost. The number of self-attention blocks was reduced for feature refinement. To prevent quality degradation, a Convolutional Block Attention Module was added, enhancing image features by using the central frame as a reference. The proposed architecture was evaluated on two datasets, achieving the best results on Tel’s dataset in terms of quality metrics. The computational cost indicated that the architecture was significantly more efficient than other Transformer-based approaches for reconstruction. The results of this research suggest that low-complexity Transformer-based architectures have great potential, with applications extending beyond HDR reconstruction to other domains.
format Article
id doaj-art-b8c4adaa7c694230aa0d171fc9766cd1
institution DOAJ
issn 1424-8220
language English
publishDate 2025-02-01
publisher MDPI AG
record_format Article
series Sensors
spelling doaj-art-b8c4adaa7c694230aa0d171fc9766cd12025-08-20T02:52:38ZengMDPI AGSensors1424-82202025-02-01255149710.3390/s25051497An Efficient and Low-Complexity Transformer-Based Deep Learning Framework for High-Dynamic-Range Image ReconstructionJosue Lopez-Cabrejos0Thuanne Paixão1Ana Beatriz Alvarez2Diodomiro Baldomero Luque3PAVIC Laboratory, University of Acre (UFAC), Rio Branco 69915-900, BrazilPAVIC Laboratory, University of Acre (UFAC), Rio Branco 69915-900, BrazilPAVIC Laboratory, University of Acre (UFAC), Rio Branco 69915-900, BrazilPAVIC Laboratory, University of Acre (UFAC), Rio Branco 69915-900, BrazilHigh-dynamic-range (HDR) image reconstruction involves creating an HDR image from multiple low-dynamic-range images as input, providing a computational solution to enhance image quality. This task presents several challenges, such as frame misalignment, overexposure, and motion, which are addressed using deep learning algorithms. In this context, various architectures with different approaches exist, such as convolutional neural networks, diffusion networks, generative adversarial networks, and Transformer-based architectures, with the latter offering the best quality but at a high computational cost. This paper proposes an HDR reconstruction architecture using a Transformer-based approach to achieve results competitive with the state of the art while reducing computational cost. The number of self-attention blocks was reduced for feature refinement. To prevent quality degradation, a Convolutional Block Attention Module was added, enhancing image features by using the central frame as a reference. The proposed architecture was evaluated on two datasets, achieving the best results on Tel’s dataset in terms of quality metrics. The computational cost indicated that the architecture was significantly more efficient than other Transformer-based approaches for reconstruction. The results of this research suggest that low-complexity Transformer-based architectures have great potential, with applications extending beyond HDR reconstruction to other domains.https://www.mdpi.com/1424-8220/25/5/1497high-dynamic-range imagingimage reconstructionchannel attentionspatial attentionTransformers
spellingShingle Josue Lopez-Cabrejos
Thuanne Paixão
Ana Beatriz Alvarez
Diodomiro Baldomero Luque
An Efficient and Low-Complexity Transformer-Based Deep Learning Framework for High-Dynamic-Range Image Reconstruction
Sensors
high-dynamic-range imaging
image reconstruction
channel attention
spatial attention
Transformers
title An Efficient and Low-Complexity Transformer-Based Deep Learning Framework for High-Dynamic-Range Image Reconstruction
title_full An Efficient and Low-Complexity Transformer-Based Deep Learning Framework for High-Dynamic-Range Image Reconstruction
title_fullStr An Efficient and Low-Complexity Transformer-Based Deep Learning Framework for High-Dynamic-Range Image Reconstruction
title_full_unstemmed An Efficient and Low-Complexity Transformer-Based Deep Learning Framework for High-Dynamic-Range Image Reconstruction
title_short An Efficient and Low-Complexity Transformer-Based Deep Learning Framework for High-Dynamic-Range Image Reconstruction
title_sort efficient and low complexity transformer based deep learning framework for high dynamic range image reconstruction
topic high-dynamic-range imaging
image reconstruction
channel attention
spatial attention
Transformers
url https://www.mdpi.com/1424-8220/25/5/1497
work_keys_str_mv AT josuelopezcabrejos anefficientandlowcomplexitytransformerbaseddeeplearningframeworkforhighdynamicrangeimagereconstruction
AT thuannepaixao anefficientandlowcomplexitytransformerbaseddeeplearningframeworkforhighdynamicrangeimagereconstruction
AT anabeatrizalvarez anefficientandlowcomplexitytransformerbaseddeeplearningframeworkforhighdynamicrangeimagereconstruction
AT diodomirobaldomeroluque anefficientandlowcomplexitytransformerbaseddeeplearningframeworkforhighdynamicrangeimagereconstruction
AT josuelopezcabrejos efficientandlowcomplexitytransformerbaseddeeplearningframeworkforhighdynamicrangeimagereconstruction
AT thuannepaixao efficientandlowcomplexitytransformerbaseddeeplearningframeworkforhighdynamicrangeimagereconstruction
AT anabeatrizalvarez efficientandlowcomplexitytransformerbaseddeeplearningframeworkforhighdynamicrangeimagereconstruction
AT diodomirobaldomeroluque efficientandlowcomplexitytransformerbaseddeeplearningframeworkforhighdynamicrangeimagereconstruction