一种利用特征预对齐的高动态范围图像重建网络

doi:10.19678/j.issn.1000-3428.0070697

计算机工程 ›› 2026, Vol. 52 ›› Issue (6): 170-178. doi: 10.19678/j.issn.1000-3428.0070697

• 计算机视觉与图形图像处理 • 上一篇下一篇

一种利用特征预对齐的高动态范围图像重建网络

谭台哲¹^,², 龚智远¹^,*(), 杨卓¹^,³

1. 广东工业大学计算机学院, 广东广州 510006
2. 河源市湾区数字经济技术发展有限公司, 广东河源 517400
3. 广东省人类运动表现科学重点实验室, 广东广州 510500

收稿日期:2024-12-10 修回日期:2025-01-13 出版日期:2026-06-15 发布日期:2025-03-14
通讯作者: 龚智远
作者简介:
谭台哲, 男, 副教授、博士, 主研方向为计算机视觉、机器学习
龚智远(通信作者), 硕士研究生
杨卓, 讲师、博士
基金资助:
国家自然科学基金(62237001); 国家自然科学基金(61907009)

A High Dynamic Range Image Reconstruction Network Using Feature Pre-Alignment

TAN Taizhe¹^,², GONG Zhiyuan¹^,*(), YANG Zhuo¹^,³

1. School of Computer, Guangdong University of Technology, Guangzhou 510006, Guangdong, China
2. Heyuan Bay Area Digital Economy Technology Development Co., Ltd., Heyuan 517400, Guangdong, China
3. Guangdong Key Laboratory of Human Sports Performance Science, Guangzhou 510500, Guangdong, China

Received:2024-12-10 Revised:2025-01-13 Online:2026-06-15 Published:2025-03-14
Contact: GONG Zhiyuan

摘要/Abstract

摘要：

基于多张不同曝光度的低动态范围(LDR)图像来重建高动态范围(HDR)图像, 是一项具有挑战性的任务, 尤其在相机和物体运动的情况下, 运动区域常常会产生伪影, 从而影响最终的重建图像质量。导致这一问题的根本原因在于多张LDR图像在内容上不对齐, 图像之间的几何差异会显著影响重建效果。为了解决该问题, 提出一种基于特征预对齐的HDR图像重建网络, 旨在通过特征的预对齐来提高HDR重建的效果。该网络由特征预对齐和HDR重建2个主要部分组成: 在特征预对齐部分, 设计一个特征预对齐网络(FPAN), 该网络能够引导输入图像的特征与参考图像对齐, 从而减少由运动引起的伪影; 在HDR重建部分, 通过选择性状态空间模型来对预对齐后的特征进行全局上下文的建模, 并通过简化的HDR恢复网络生成最终的HDR图像。为了评估所提网络的性能, 在2个数据集上进行广泛的实验, 结果显示, 该网络在多个客观评估指标上优于对比方法, 在主观视觉上也呈现出良好的效果, 且具有一定的泛化能力。

关键词: 高动态范围图像, 图像重建, 特征预对齐, 选择性状态空间模型, 运动伪影

Abstract:

Reconstructing High Dynamic Range (HDR) images from multiple Low Dynamic Range (LDR) images with varying exposures is challenging, especially in scenarios involving camera and object motion, where motion regions often introduce artifacts, thereby affecting the quality of the final reconstructed image. This issue primarily arises from the misalignment of content among multiple LDR images, where geometric differences between images significantly impact the reconstruction outcome. To address this issue, we propose an HDR image reconstruction network based on feature pre-alignment for enhancing HDR reconstruction quality. This network has two stages: feature pre-alignment and HDR reconstruction. In the feature pre-alignment stage, a Feature Pre-Alignment Network (FPAN) guides the alignment of features from input images with those from the reference image, thereby reducing artifacts caused by motion. In HDR reconstruction stage, a selective state space model is employed for modeling the global context of the pre-aligned features, and a simplified HDR restoration network generates the final HDR image. Extensive experiments are conducted on two datasets to evaluate the performance of the proposed network. The results show that the proposed network outperforms comparative methods across multiple objective evaluation metrics, exhibits satisfactory subjective visual effects, and demonstrates certain generalization capabilities.

Key words: High Dynamic Range (HDR) images, image reconstruction, feature pre-alignment, selective state space model, motion artifacts

谭台哲, 龚智远, 杨卓. 一种利用特征预对齐的高动态范围图像重建网络[J]. 计算机工程, 2026, 52(6): 170-178.

TAN Taizhe, GONG Zhiyuan, YANG Zhuo. A High Dynamic Range Image Reconstruction Network Using Feature Pre-Alignment[J]. Computer Engineering, 2026, 52(6): 170-178.

https://www.ecice06.com/CN/Y2026/V52/I6/170

图/表 10

图1 本文网络整体架构

Fig.1 The overall architecture of the network in this paper

图2 改进的Mamba结构和二维选择性扫描模块

Fig.2 Improved Mamba structure and 2D selective scanning module

图3 SG和SCA的结构

Fig.3 SG and SCA structures

图4 文献[7]测试数据集中的“停车”场景下不同方法的HDR重建结果比较

Fig.4 Comparison of HDR reconstruction results using different methods in the ″parking″ scenario from the test dataset in reference [7]

图5 文献[26]数据集上的定性结果

Fig.5 Qualitative results on the dataset from reference [26]

图6 消融实验定性结果

Fig.6 Qualitative results of ablation experiment

参考文献 34

1	HU J, GALLO O, PULLI K, et al. HDR deghosting: how to deal with saturation?[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Washington D.C., USA: IEEE Press, 2013: 1163-1170.
2	KANG S B , UYTTENDAELE M , WINDER S , et al. High dynamic range video. ACM Transactions on Graphics, 2003, 22 (3): 319- 325. doi: 10.1145/882262.882270
3	ZIMMER H , BRUHN A , WEICKERT J . Freehand HDR imaging of moving scenes with simultaneous resolution enhancement. Computer Graphics Forum, 2011, 30 (2): 405- 414. doi: 10.1111/j.1467-8659.2011.01870.x
4	GALLO O, GELFANDZ N, CHEN W C, et al. Artifact-free high dynamic range imaging[C]//Proceedings of the IEEE International Conference on Computational Photography (ICCP). Washington D.C., USA: IEEE Press, 2010: 1-7.
5	JACOBS K , LOSCOS C , WARD G . Automatic high-dynamic range image generation for dynamic scenes. IEEE Computer Graphics and Applications, 2008, 28 (2): 84- 93. doi: 10.1109/MCG.2008.23
6	PECE F, KAUTZ J. Bitmap movement detection: HDR for dynamic scenes[C]//Proceedings of the Conference on Visual Media Production. Washington D.C., USA: IEEE Press, 2011: 1-8.
7	KALANTARI N K , RAMAMOORTHI R . Deep high dynamic range imaging of dynamic scenes. ACM Transactions on Graphics, 2017, 36 (4): 1- 12.
8	PRABHAKAR K R, AGRAWAL S, SINGH D K, et al. Towards practical and efficient high-resolution HDR deghosting with CNN[EB/OL]. [2024-08-05]. https://link.springer.com/chapter/10.1007/978-3-030-58589-1_30.
9	WU S Z, XU J R, TAI Y W, et al. Deep high dynamic range imaging with large foreground motions[EB/OL]. [2024-08-05]. https://arxiv.org/abs/1711.08937.
10	CHUNG H, CHO N I. High dynamic range imaging of dynamic scenes with saturation compensation but without explicit motion compensation[C]// Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV). Washington D.C., USA: IEEE Press, 2022: 61-71.
11	LIU Z, LIN W J, LI X P, et al. ADNet: attention-guided deformable convolutional network for high dynamic range imaging[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW). Washington D.C., USA: IEEE Press, 2021: 463-470.
12	TAN X , CHEN H A , XU K , et al. High dynamic range imaging for dynamic scenes with large-scale motions and severe saturation. IEEE Transactions on Instrumentation and Measurement, 2022, 71, 5003415.
13	ALEXEY D. An image is worth 16×16 words: Transformers for image recognition at scale[EB/OL]. [2024-08-05]. https://arxiv.org/abs/2010.11929.
14	田永林, 王雨桐, 王建功, 等. 视觉Transformer研究的关键问题: 现状及展望. 自动化学报, 2022, 48 (4): 957- 979.
	TIAN Y L , WANG Y T , WANG J G , et al. Key problems and progress of vision Transformer: the state of the art and prospects. Acta Automatica Sinica, 2022, 48 (4): 957- 979.
15	彭斌, 白静, 李文静, 等. 面向图像分类的视觉Transformer研究进展. 计算机科学与探索, 2024, 18 (2): 320- 344.
	PENG B , BAI J , LI W J , et al. Survey on visual Transformer for image classification. Journal of Frontiers of Computer Science and Technology, 2024, 18 (2): 320- 344.
16	LIU Z, WANG Y L, ZENG B, et al. Ghost-free high dynamic range imaging with context-aware Transformer[EB/OL]. [2024-08-05]. https://arxiv.org/abs/2208.05114.
17	YAN Q S, CHEN W Y, ZHANG S, et al. A unified HDR imaging method with pixel and patch level[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Washington D.C., USA: IEEE Press, 2023: 22211-22220.
18	CHEN R F , ZHENG B L , ZHANG H , et al. Improving dynamic HDR imaging with fusion Transformer. Proceedings of the AAAI Conference on Artificial Intelligence, 2023, 37 (1): 340- 349. doi: 10.1609/aaai.v37i1.25107
19	GU A, DAO T. Mamba: linear-time sequence modeling with selective state spaces[EB/OL]. [2024-08-05]. https://arxiv.org/abs/2312.00752.
20	EILERTSEN G , KRONANDER J , DENES G , et al. HDR image reconstruction from a single exposure using deep CNNs. ACM Transactions on Graphics, 2017, 36 (6): 1- 15.
21	贺姗, 蔺素珍, 王彦博, 等. 基于特征融合的多波段图像描述生成方法. 计算机工程, 2024, 50 (6): 236- 244. doi: 10.19678/j.issn.1000-3428.0067302
	HE S , LIN S Z , WANG Y B , et al. Multi-band image caption generation method based on feature fusion. Computer Engineering, 2024, 50 (6): 236- 244. doi: 10.19678/j.issn.1000-3428.0067302
22	GUO H, LI J M, DAI T, et al. MambaIR: a simple baseline for image restoration with state-space model[EB/OL]. [2024-08-05]. https://arxiv.org/abs/2402.15648.
23	ZHU L H, LIAO B C, ZHANG Q, et al. Vision Mamba: efficient visual representation learning with bidirectional state space model[EB/OL]. [2024-08-05]. https://arxiv.org/html/2401.09417v3.
24	CHEN S, ATAPOUR A A, ZHANG H Z, et al. MxT: Mamba x Transformer for image inpainting[EB/OL]. [2024-08-05]. https://arxiv.org/abs/2407.16126.
25	CHEN L Y, CHU X J, ZHANG X Y, et al. Simple baselines for image restoration[EB/OL]. [2024-08-05]. https://arxiv.org/abs/2204.04676.
26	SEN P , KALANTARI N K , YAESOUBI M , et al. Robust patch-based HDR reconstruction of dynamic scenes. ACM Transactions on Graphics, 2012, 31 (6): 1- 11.
27	MANTIUK R , KIM K J , REMPEL A G , et al. HDR-VDP-2: a calibrated visual metric for visibility and quality predictions in all luminance conditions. ACM Transactions on Graphics, 2011, 30 (4): 1- 14.
28	EILERTSEN G, HAJISHARIF S, HANJI P, et al. How to cheat with metrics in single-image HDR reconstruction[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision Workshops (ICCVW). Washington D.C., USA: IEEE Press, 2021: 3981-3990.
29	杨珍妹, 李华锋, 张亚飞. 面向高动态范围成像的内容恢复和鬼影抑制网络. 模式识别与人工智能, 2024, 37 (4): 313- 327.
	YANG Z M , LI H F , ZHANG Y F . Missing content restoration and ghosting suppression network for high dynamic range imaging. Pattern Recognition and Artificial Intelligence, 2024, 37 (4): 313- 327.
30	GU K , WANG S Q , ZHAI G T , et al. Blind quality assessment of tone-mapped images via analysis of information, naturalness, and structure. IEEE Transactions on Multimedia, 2016, 18 (3): 432- 443. doi: 10.1109/TMM.2016.2518868
31	FANG Y M , ZHU H W , MA K D , et al. Perceptual evaluation for multi-exposure image fusion of dynamic scenes. IEEE Transactions on Image Processing, 2020, 29, 1127- 1138. doi: 10.1109/TIP.2019.2940678
32	YAN Q S, GONG D, SHI Q F, et al. Attention-guided network for ghost-free high dynamic range imaging[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Washington D.C., USA: IEEE Press, 2020: 1751-1760.
33	TANG L F , HUANG H , ZHANG Y F , et al. Structure-embedded ghosting artifact suppression network for high dynamic range image reconstruction. Knowledge-Based Systems, 2023, 263, 110278. doi: 10.1016/j.knosys.2023.110278
34	NIU Y Z , WU J B , LIU W X , et al. HDR-GAN: HDR image reconstruction from multi-exposed LDR images with large motions. IEEE Transactions on Image Processing, 2021, 30, 3885- 3896. doi: 10.1109/TIP.2021.3064433

[1]	曹坪, 杨怀志, 薄一军, 尤嘉, 张淳杰, 李丹勇. 面向低质量裂缝图像的多知识蒸馏分类[J]. 计算机工程, 2023, 49(7): 204-213.
[2]	甄诚, 杨永胜, 李元祥, 钟娟娟. 基于多尺度生成对抗网络的大气湍流图像复原[J]. 计算机工程, 2021, 47(11): 227-233.
[3]	陈宇,李洪宇,夏宗基. 基于修正隐式Landweber方法的ECT图像重建算法[J]. 计算机工程, 2018, 44(1): 268-273.
[4]	司凯,孙丰荣,宋尚玲,秦峰. 基于折射角的稀疏投影角度相衬CT 图像重建[J]. 计算机工程, 2015, 41(3): 262-268.
[5]	陈洪磊, 贺建峰, 刘俊卿. 基于二维检索的投影矩阵算法[J]. 计算机工程, 2013, 39(2): 229-232.
[6]	曾宪华, 段文强. 基于近邻非负线性组合的高分辨率图像重建[J]. 计算机工程, 2012, 38(22): 211-215.
[7]	王欢, 王永革. 基于L1/2正则化的超分辨率图像重建算法[J]. 计算机工程, 2012, 38(20): 191-194.
[8]	廉飞宇, 李青, 秦瑶. 储粮水分含量异常区域的雷达层析成像检测[J]. 计算机工程, 2012, 38(18): 198-202.
[9]	侯金曼, 何宁, 吕科. 基于压缩感知的图像快速重建方法[J]. 计算机工程, 2011, 37(19): 215-217,226.
[10]	马自萍, 康宝生. 贝塞尔-傅里叶矩在图像检索中的应用研究[J]. 计算机工程, 2011, 37(17): 194-196.
[11]	孙琰玥, 何小海, 陈为龙. 小波局部适应插值的图像超分辨率重建[J]. 计算机工程, 2010, 36(13): 183-185.
[12]	牛小明;曾理;刘宝东;邹晓兵;. 偏心螺旋锥束CT的Katsevich修正重建算法[J]. 计算机工程, 2009, 35(5): 215-217,.
[13]	刘鹏;刘定生;李国庆. 基于矩阵秩估计偏移量的频域超分辨率重建[J]. 计算机工程, 2009, 35(15): 29-31,3.
[14]	樊瑶;赵祥模;褚燕利;党乐. 基于预条件共轭梯度法的混凝土层析成像[J]. 计算机工程, 2008, 34(23): 258-260.
[15]	马存宝;周伟;张潇;左云霞. 基于改进二维图像重建算法的工业CT系统[J]. 计算机工程, 2007, 33(17): 219-221，.

选择文件类型/文献管理软件名称

选择包含的内容

一种利用特征预对齐的高动态范围图像重建网络

A High Dynamic Range Image Reconstruction Network Using Feature Pre-Alignment

RichHTML

PDF

可视化

被引次数

摘要/Abstract

引用本文

使用本文

图/表 10

参考文献 34

相关文章 15

编辑推荐

Metrics

本文评价

模态框（Modal）标题

选择文件类型/文献管理软件名称

选择包含的内容

一种利用特征预对齐的高动态范围图像重建网络

A High Dynamic Range Image Reconstruction Network Using Feature Pre-Alignment

RichHTML

PDF

可视化

被引次数

摘要/Abstract

引用本文

使用本文

图/表 10

参考文献 34

相关文章 15

编辑推荐

Metrics

本文评价