Generative Adversarial Network-Based Lightweight High-Dynamic-Range Image Reconstruction Model

The generation of High-Dynamic-Range (HDR) images is essential for capturing details at various brightness levels, but current reconstruction methods, using deep learning techniques, often require significant computational resources, limiting their applicability on devices with moderate resources. I...

Full description

Saved in:
Bibliographic Details
Main Authors: Gustavo de Souza Ferreti, Thuanne Paixão, Ana Beatriz Alvarez
Format: Article
Language:English
Published: MDPI AG 2025-04-01
Series:Applied Sciences
Subjects:
Online Access:https://www.mdpi.com/2076-3417/15/9/4801
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850278567948058624
author Gustavo de Souza Ferreti
Thuanne Paixão
Ana Beatriz Alvarez
author_facet Gustavo de Souza Ferreti
Thuanne Paixão
Ana Beatriz Alvarez
author_sort Gustavo de Souza Ferreti
collection DOAJ
description The generation of High-Dynamic-Range (HDR) images is essential for capturing details at various brightness levels, but current reconstruction methods, using deep learning techniques, often require significant computational resources, limiting their applicability on devices with moderate resources. In this context, this paper presents a lightweight architecture for reconstructing HDR images from three Low-Dynamic-Range inputs. The proposed model is based on Generative Adversarial Networks and replaces traditional convolutions with depthwise separable convolutions, reducing the number of parameters while maintaining high visual quality and minimizing luminance artifacts. The evaluation of the proposal is conducted through quantitative, qualitative, and computational cost analyses based on the number of parameters and FLOPs. Regarding the qualitative analysis, a comparison between the models was performed using samples that present reconstruction challenges. The proposed model achieves a PSNR-<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mi>μ</mi></semantics></math></inline-formula> of 43.51 dB and SSIM-<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mi>μ</mi></semantics></math></inline-formula> of 0.9917, achieving competitive quality metrics comparable to HDR-GAN while reducing the computational cost by 6× in FLOPs and 7× in the number of parameters, using approximately half the GPU memory consumption, demonstrating an effective balance between visual fidelity and efficiency.
format Article
id doaj-art-704f087275b84dbf9260fccc7d9707c2
institution OA Journals
issn 2076-3417
language English
publishDate 2025-04-01
publisher MDPI AG
record_format Article
series Applied Sciences
spelling doaj-art-704f087275b84dbf9260fccc7d9707c22025-08-20T01:49:27ZengMDPI AGApplied Sciences2076-34172025-04-01159480110.3390/app15094801Generative Adversarial Network-Based Lightweight High-Dynamic-Range Image Reconstruction ModelGustavo de Souza Ferreti0Thuanne Paixão1Ana Beatriz Alvarez2PAVIC 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, BrazilThe generation of High-Dynamic-Range (HDR) images is essential for capturing details at various brightness levels, but current reconstruction methods, using deep learning techniques, often require significant computational resources, limiting their applicability on devices with moderate resources. In this context, this paper presents a lightweight architecture for reconstructing HDR images from three Low-Dynamic-Range inputs. The proposed model is based on Generative Adversarial Networks and replaces traditional convolutions with depthwise separable convolutions, reducing the number of parameters while maintaining high visual quality and minimizing luminance artifacts. The evaluation of the proposal is conducted through quantitative, qualitative, and computational cost analyses based on the number of parameters and FLOPs. Regarding the qualitative analysis, a comparison between the models was performed using samples that present reconstruction challenges. The proposed model achieves a PSNR-<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mi>μ</mi></semantics></math></inline-formula> of 43.51 dB and SSIM-<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mi>μ</mi></semantics></math></inline-formula> of 0.9917, achieving competitive quality metrics comparable to HDR-GAN while reducing the computational cost by 6× in FLOPs and 7× in the number of parameters, using approximately half the GPU memory consumption, demonstrating an effective balance between visual fidelity and efficiency.https://www.mdpi.com/2076-3417/15/9/4801HDR reconstructionmulti-exposed imaginglarge motiondepthwise separable convolutiongenerative adversarial network
spellingShingle Gustavo de Souza Ferreti
Thuanne Paixão
Ana Beatriz Alvarez
Generative Adversarial Network-Based Lightweight High-Dynamic-Range Image Reconstruction Model
Applied Sciences
HDR reconstruction
multi-exposed imaging
large motion
depthwise separable convolution
generative adversarial network
title Generative Adversarial Network-Based Lightweight High-Dynamic-Range Image Reconstruction Model
title_full Generative Adversarial Network-Based Lightweight High-Dynamic-Range Image Reconstruction Model
title_fullStr Generative Adversarial Network-Based Lightweight High-Dynamic-Range Image Reconstruction Model
title_full_unstemmed Generative Adversarial Network-Based Lightweight High-Dynamic-Range Image Reconstruction Model
title_short Generative Adversarial Network-Based Lightweight High-Dynamic-Range Image Reconstruction Model
title_sort generative adversarial network based lightweight high dynamic range image reconstruction model
topic HDR reconstruction
multi-exposed imaging
large motion
depthwise separable convolution
generative adversarial network
url https://www.mdpi.com/2076-3417/15/9/4801
work_keys_str_mv AT gustavodesouzaferreti generativeadversarialnetworkbasedlightweighthighdynamicrangeimagereconstructionmodel
AT thuannepaixao generativeadversarialnetworkbasedlightweighthighdynamicrangeimagereconstructionmodel
AT anabeatrizalvarez generativeadversarialnetworkbasedlightweighthighdynamicrangeimagereconstructionmodel