采用上下文双自相关注意力的偏振图像超分辨率重建网络

陈广秋; 张赛; 刘枫铭; 贯成才; 刘国峰; 段锦; 姜洪艳

doi:null

您当前的位置：

首页 >

文章列表页 >

采用上下文双自相关注意力的偏振图像超分辨率重建网络

更新时间：2026-07-21

- 采用上下文双自相关注意力的偏振图像超分辨率重建网络
- Polarization Image Super-Resolution Reconstruction Network Using Contextual Dual Self-Correlation Attention
- 西安交通大学学报 2026年
- 作者机构：
  
  1.长春理工大学电子信息工程学院，130022，长春
  2.长春市吉海测控技术有限责任公司，130012，长春
- 作者简介：
- 基金信息：
- DOI：
  中图分类号： TP391
- 收稿：2026-04-02，
  
  修回：2026-07-21，
  
  录用：2026-07-21，
- 稿件说明：
移动端阅览
陈广秋, 张赛, 刘枫铭, 等. 采用上下文双自相关注意力的偏振图像超分辨率重建网络[J/OL]. 西安交通大学学报, 2026.

CHEN Guangqiu, ZHANG Sai, LIU Fengming, et al. Polarization Image Super-Resolution Reconstruction Network Using Contextual Dual Self-Correlation Attention[J/OL]. JOURNAL OF XI’AN JIAOTONG UNIVERSITY, 2026.
陈广秋, 张赛, 刘枫铭, 等. 采用上下文双自相关注意力的偏振图像超分辨率重建网络[J/OL]. 西安交通大学学报, 2026. DOI：

CHEN Guangqiu, ZHANG Sai, LIU Fengming, et al. Polarization Image Super-Resolution Reconstruction Network Using Contextual Dual Self-Correlation Attention[J/OL]. JOURNAL OF XI’AN JIAOTONG UNIVERSITY, 2026. DOI：

摘要

针对彩色分焦平面偏振图像全分辨率重建中，存在的颜色失真、边缘退化和偏振参数不一致问题，提出一种基于上下文双自相关注意力的偏振图像超分辨率重建网络。该网络由双路特征提取模块、上下文双自相关注意力Transformer和偏振特征重建模块组成。其中，双路特征提取模块用于增强浅层结构与纹理特征；上下文双自相关注意力Transformer通过上下文门控融合、位置感知自相关注意力机制和局部深度自相关注意力机制，建立不同偏振方向之间的空间结构关系与通道相关性；偏振特征重建模块通过特征精炼和像素重排恢复全分辨率彩色偏振图像。研究结果表明：在SunPol-Real、OL和Monno数据集上，所提网络对低分辨图像进行超分辨重建，0°、45°、90°和135°这4个方向图像的峰值信噪比的平均值分别达到38.16、39.39和40.02 dB，结构相似性的平均值分别达到0.967 6、0.965 8和0.981 5，从而验证了该网络能够改善边缘模糊、纹理缺失和偏振信息不一致的问题。该研究为彩色分焦平面偏振图像的超分辨率重建提供了一种有效的技术路径。

Abstract

To address the problems of color distortion

edge degradation

and inconsistency in polarization parameters during the full-resolution reconstruction of color division-of-focal-plane (DoFP) polarization images

a polarization image super-resolution reconstruction network based on contextual dual autocorrelation attention is proposed. The network consists of a dual-branch feature extraction module

a contextual dual-autocorrelation attention Transformer

and a polarization feature reconstruction module. Specifically

the dual-branch feature extraction module is employed to enhance shallow structural and texture features. The contextual dual-autocorrelation attention Transformer establishes spatial structural relationships and channel correlations among different polarization orientations through contextual gated fusion

position-aware autocorrelation attention

and local depth-wise autocorrelation attention mechanisms. The polarization feature reconstruction module restores full-resolution color polarization images through feature refinement and pixel rearrangement. Experimental results show that

on the SunPol-Real

and Monno datasets

the proposed network achieves average peak signal-to-noise ratio values of 38.16

39.39

and 40.02 dB

respectively

and average structural similarity index values of 0.967 6

0.965 8

and 0.981 5

respectively

across the reconstructed images at the 0°

45°

90°

and 135° polarization orientations. These results demonstrate that the proposed network can effectively alleviate edge blurring and texture loss while improving the consistency of polarization information. This study provides an effective technical approach for the super-resolution reconstruction of color DoFP polarization images.

关键词

Keywords

references

YU Dabing , LI Qingwu , ZHANG Zhiliang , et al . Color polarization image super-resolution reconstruction via a cross-branch supervised learning strategy [J ] . Optics and Lasers in Engineering , 2023 , 165 : 107469 .

ZHANG Jingjing , BAO Kangsheng , ZHANG Xin , et al . Conditional generative adversarial defogging algorithm based on polarization characteristics [C ] // Proceedings of the 2021 33rd Chinese Control and Decision Conference (CCDC) . Piscataway, NJ, USA : IEEE , 2021 : 2148 – 2155 .

吴洪伍 , 盖绍彦 , 达飞鹏 . 采用感受野优化与渐进特征融合的图像超分辨率算法 [J ] . 西安交通大学学报 , 2025 , 59 ( 01 ): 136 - 147 .

WU Hongwu , GAI Shaoyan , DA Feipeng . Image super-resolution algorithm based on receptive field optimization and progressive feature fusion [J ] . Journal of Xi'an Jiaotong University , 2025 , 59 ( 1 ): 136 – 147 .

SU Yiming , HAO Chunhui , CHEN Wen , et al . Universal polarization demosaicking using Newton polynomial and residual interpolation [J ] . IEEE Sensors Journal , 2025 , 25 ( 7 ): 1 1508– 11520 .

SI Weijian , WANG Yan , ZHANG Chunjie . Three-parallel co-prime polarization sensitive array for 2-D DOA and polarization estimation via sparse representation [J ] . IEEE Access , 2019 , 7 : 1 5404– 15413 .

CAO Yin , ZHANG Heng , FAN Rong , et al . Underwater image restoration based on color polarization imaging and deep learning [C ] // Proceedings of the 2025 6th International Conference on Computer Vision, Image and Deep Learning (CVIDL) . Piscataway, NJ, USA : IEEE , 2025 : 612 – 616 .

ZHANG Junchao , SHAO Jianbo , LUO Haibo , et al . Learning a convolutional demosaicing network for microgrid polarimeter imagery [J ] . Optics Letters , 2018 , 43 ( 18 ): 4534 – 4537 .

WEN Sijia , ZHENG Yinqiang , LU Feng , et al . Convolutional demosaicing network for joint chromatic and polarimetric imagery [J ] . Optics Letters , 2019 , 44 ( 22 ): 5646 – 5649 .

SUN Yuanyuan , ZHANG Junchao , LIANG Rongguang . Color polarization demosaicking by a convolutional neural network [J ] . Optics Letters , 2021 , 46 ( 17 ): 4338 – 4341 .

GUAN Chao , LI Jiangtao , TIAN Yaping , et al . Polarization-enhanced super-resolution imaging reconstruction based on a denoising diffusion probabilistic model [J ] . Optics Express , 2025 , 33 ( 19 ): 3 9238– 39248 .

LI Shuoshi , ZHOU Yuan , KUNG S Y . PSRNet: A progressive self-refine network for lightweight optical remote sensing image dehazing [J ] . IEEE Transactions on Geoscience and Remote Sensing , 2024 , 62 : 1 – 13 , Art. no. 5648413 .

ZHOU Shuangfan , ZHOU Chu , LYU Youwei , et al . PIDSR: Complementary polarized image demosaicing and super-resolution [C ] // Proceedings of the 2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) . Piscataway, NJ, USA : IEEE , 2025 : 16081 – 16090 .

WANG Wenxiao , CHEN Wei , QIU Qibo , et al . CrossFormer++: A versatile vision transformer hinging on cross-scale attention [J ] . IEEE Transactions on Pattern Analysis and Machine Intelligence , 2024 , 46 ( 5 ): 3123 – 3136 .

陈广秋 , 刘枫铭 , 段锦 , 等 . 基于行列栅格感知Transformer的车道线检测方法 [J ] . 光学精密工程 , 2026 , 34 ( 6 ): 953 – 972 .

CHEN Guangqiu , LIU Fengming , DUAN Jin , et al . Row-column grid-aware Transformer based lane detection method [J ] . Optics and Precision Engineering , 2026 , 34 ( 6 ): 953 – 972 .

左龙 , 张鹏 , 荆树旭 , 等 . 用于图像超分辨率重建的双通道残差网络 [J ] . 西安交通大学学报 , 2022 , 56 ( 1 ): 158 – 164 .

ZUO Long , ZHANG Peng , JING Shuxu , et al . Dual-channel residual network for image super-resolution reconstruction [J ] . Journal of Xi’an Jiaotong University , 2022 , 56 ( 1 ): 158 – 164 .

GIL J J , OSSIKOVSKI R . Polarized light and the Mueller matrix approach [M ] . Boca Raton, FL, USA : CRC Press , 2022 .

HU Haofeng , YANG Shiyao , LI Xiaobo , et al . Polarized image super-resolution via a deep convolutional neural network [J ] . Optics Express , 2023 , 31 ( 5 ): 8535 – 8547 .

MORIMATSU M , MONNO Y , TANAKA M , et al . Monochrome and color polarization demosaicking based on intensity-guided residual interpolation [J ] . IEEE Sensors Journal , 2021 , 21 ( 23 ): 2 6985– 26996 .

MORIMATSU M , MONNO Y , TANAKA M , et al . Monochrome and color polarization demosaicking using edge-aware residual interpolation [C ] // Proceedings of the 2020 IEEE International Conference on Image Processing (ICIP) . Piscataway, NJ, USA : IEEE , 2020 : 2571 – 2575 .

WU Rongyuan , ZHAO Yongqiang , LI Ning , et al . Polarization image demosaicking using polarization channel difference prior [J ] . Optics Express , 2021 , 29 ( 14 ): 2 2066– 22079 .

LUO Yidong , ZHANG Junchao , SHAO Jianbo , et al . Learning a non-locally regularized convolutional sparse representation for joint chromatic and polarimetric demosaicking [J ] . IEEE Transactions on Image Processing , 2024 , 33 : 5029 – 5044 .

ZHANG Xiang , ZHANG Yulun , YU Fisher . HiT-SR: Hierarchical transformer for efficient image super-resolution [C ] // Computer Vision—ECCV 2024 . Cham, Switzerland : Springer , 2024 : 483 – 500 .

徐胜军 , 邓博文 , 史亚 , 等 . 一种编解码结构的车牌图像超分辨率网络 [J ] . 西安交通大学学报 , 2022 , 56 ( 10 ): 101 – 110 .

XU Shengjun , DENG Bowen , SHI Ya , et al . An encoder-decoder structured network for license plate image super-resolution [J ] . Journal of Xi’an Jiaotong University , 2022 , 56 ( 10 ): 101 – 110 .

高峰 , 汲胜昌 , 郭洁 , 等 . 采用对比学习的多阶段Transformer图像去雾方法 [J ] . 西安交通大学学报 , 2023 , 57 ( 1 ): 195 – 210 .

GAO Feng , JI Shengchang , GUO Jie , et al . A multi-stage Transformer network for image dehazing based on contrastive learning [J ] . Journal of Xi’an Jiaotong University , 2023 , 57 ( 1 ): 195 – 210 .

LIU Yibin , WANG Shengbin Luo , WU Guoqing , et al . Microwave image super-resolution using polarization [C ] // Proceedings of the 2025 10th International Conference on Signal and Image Processing (ICSIP) . Piscataway, NJ, USA : IEEE , 2025 : 258 – 262 .

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

自适应估计模糊参数的最大后验概率超分辨率复原算法

全局重建和位置块残差补偿的人脸图像超分辨率算法

利用位置字典对的人脸图像超分辨率方法

人脸超分辨率重建中投影空间的选择方法

多尺度区域注意力InfoGAN车牌识别网络