复杂运动场景下的视频图像湍流抑制  

Turbulence mitigation method for complex dynamic scene videos

作　　者：景元萍 胡银记 芦碧波[3] JING Yuanping;HU Yinji;LU Bibo(Department of Mathematics and Physics,Luoyang Institute of Science and Technology,Luoyang 471000,China;Luoyang Institute of Electro‑Optical Equipment,Aviation Industry Corporation of China,Luoyang 471000,China;College of Computer Science and Technology,Henan University of Technology,Jiaozuo 454000,China)

机构地区：[1]洛阳理工学院数学与物理教学部,河南洛阳471000 [2]中国航空工业集团公司洛阳电光设备研究所,河南洛阳471000 [3]河南理工大学计算机科学与技术学院,河南焦作454000

出　　处：《信阳师范大学学报(自然科学版)》2025年第2期167-173,共7页Journal of Xinyang Normal University(Natural Science Edition)

基　　金：国家自然科学基金项目(42272178);航空科学基金项目(2022Z015013002);2022年度河南省重点研发与推广专项(科技攻关)项目(222102210131)。

摘　　要：针对大多数湍流抑制算法在复杂运动场景下复原效果无法满足需求的现状,提出了一种可适用于运动场景与运动目标的快速高清视频湍流抑制技术。首先,采用基于灰度投影的多级搜索匹配方法完成场景的快速配准;然后,结合结构一致性与清晰度判断,完成运动目标与背景自适应融合,实现湍流扭曲的去除与运动区域的保持;最后,采用基于梯度高斯分布先验的空域复原方法,将复杂的频域复原问题转化为简单的空域滤波,提升图像清晰度的同时,降低计算复杂度。实验结果表明:无论是定量指标还是直接视觉感知结果,均较同类方法具有显著的优势,并且计算复杂度低,具有良好的嵌入式应用前景。Aiming at the current situation where most of turbulence suppression algorithms cannot meet the restoration requirements in complex motion scenes,a fast high‑definition video turbulence suppression technology applicable to motion scenes and moving targets was proposed.Firstly,a multi‑level search matching method based on grayscale projection was used to achieve fast scene registration.Then,by combining structural consistency and clarity judgment,the adaptive fusion of moving targets and backgrounds was achieved,achieving the removal of turbulence distortion and the preservation of moving regions.Finally,a spatial domain restoration method based on gradient Gaussian distribution prior was adopted to transform complex frequency domain restoration problems into simple spatial filtering,improving image clarity while reducing computational complexity.Experimental results showed that both quantitative indicators and direct visual perception results had significant advantages over similar methods,and had low computational complexity,making them promising for embedded applications.

关 键 词：大气湍流 运动场景 图像复原 视频图像 湍流抑制 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]