航空学报 > 2025, Vol. 46 Issue (23): 631652-631652   doi: 10.7527/S1000-6893.2025.31652

干扰环境下无人机多源感知专栏

双向映射与物理负相关的航空图像去雾算法

姚兆汝1, 孙航1,2(), 任东1,2, 刘莉1,2, 万俊3   

  1. 1.三峡大学 计算机与信息学院,宜昌 443002
    2.三峡大学 水电工程智能视觉监测湖北省重点实验室,宜昌 443002
    3.中南财经政法大学 信息工程学院,武汉 430073
  • 收稿日期:2024-12-11 修回日期:2025-01-10 接受日期:2025-02-21 出版日期:2025-03-19 发布日期:2025-03-19
  • 通讯作者: 孙航 E-mail:sunhang@ctgu.edu.cn
  • 基金资助:
    国家自然科学基金(62576192);国家自然科学基金(42401404)

Bidirectional mapping and negative correlation algorithm for aerial image dehazing

Zhaoru YAO1, Hang SUN1,2(), Dong REN1,2, Li LIU1,2, Jun WAN3   

  1. 1.College of Computer and Information Technology,China Three Gorges University,Yichang 443002,China
    2.Hubei Key Laboratory of Intelligent Vision Based Monitoring for Hydroelectric Engineering,China Three Gorges University,Yichang 443002,China
    3.School of Information Engineering,Zhongnan University of Economics and Law,Wuhan 430073,China
  • Received:2024-12-11 Revised:2025-01-10 Accepted:2025-02-21 Online:2025-03-19 Published:2025-03-19
  • Contact: Hang SUN E-mail:sunhang@ctgu.edu.cn
  • Supported by:
    National Natural Science Foundation of China(62576192)

摘要:

近年来基于深度神经网络的航空图像去雾技术取得了显著进展。然而,大多数方法仅学习从雾霾图像到清晰图像的单向映射,未充分利用雾霾生成过程中的潜在监督信息,导致去雾性能下降。此外,现有在特征空间引入大气散射模型的方法忽略了传输图与大气光之间的物理负相关性,这对学习更具代表性的特征至关重要。为此,提出了一种双向映射与物理负相关的航空图像去雾算法。具体来说,构建了双向映射多阶段框架,在共享参数的网络中协同建模雾霾去除与生成过程,有效探索了雾霾形成机制,从而提升去雾性能。设计了大气光引导的物理负相关感知模块,通过负相关引导机制利用大气光信息优化传输图特征估计。在公开航空雾霾数据集上的实验结果表明,所提方法在客观评价指标和视觉效果方面均优于当前最先进的7种去雾算法。

关键词: 航空图像去雾, 双向映射, 多阶段, 负相关, 大气散射模型

Abstract:

Recently, deep neural networks based aerial image dehazing techniques have achieved remarkable performance. However, most methods only learn a one-way mapping from hazy images to clear images, which fail to fully exploit the latent supervision information of haze generation, leading to dehazing performance degradation. Moreover, existing methods that incorporate the atmospheric scattering model in the feature space overlook the exploration of the negative correlation between the transmission map and atmospheric light, which is crucial for leaning more representative features. To address these issues, we propose a bidirectional mapping physics-aware negative correlation algorithm for aerial image dehazing. Specifically, the proposed bidirectional mapping multi-stage framework synchronously models haze removal and generation processes in a parameter-sharing network, effectively exploring the haze formation mechanism to enhance dehazing performance. Furthermore, an atmospheric light-guided physics-aware negative correlation module is designed, which optimizes transmission map estimation by leveraging a negative correlation guidance mechanism utilizing atmospheric light information. Experimental results on public datasets demonstrate that the proposed dehazing algorithm outperforms seven state-of-the art haze removal methods in both objective evaluation and visual quality.

Key words: aerial image dehazing, bidirectional mapping, multi-stage, negative correlation, atmospheric scattering model

中图分类号: