基于方向和结构特征的遥感图像条带噪声分离方法  

Stripe Separation Method for Remote Sensing Images Based on Directional and Structural Feature

作　　者：孔祥阳[1,2] 彭群聂 徐保根 KONG Xiangyang;PENG Qunnie;XU Baogen(School of Automation Northwestern Polytechnical University,Xi'an 710072 China;Ministry of Basic Education Sichuan Engineering Technical College,Deyang 618000 China;Science and Technology on Electro-Optic Control Laboratory,Luoyang 471000 China;School of Science East China Jiaotong University,Nanchang 330013 China)

机构地区：[1]西北工业大学自动化学院,西安710072 [2]四川工程职业技术学院基础教学部,四川德阳618000 [3]光电控制技术重点实验室,河南洛阳471000 [4]华东交通大学理学院,南昌330013

出　　处：《电光与控制》2020年第1期6-11,共6页Electronics Optics & Control

基　　金：国家自然科学基金(11361024)

摘　　要：针对去除遥感图像条带噪声时易造成模糊或细节丢失等问题,分析研究干净图像及条带噪声的方向和结构特征,提出了遥感图像条带噪声分离方法。对于条带噪声,其稀疏结构特征表征为条带的组稀疏性和平滑性,结合条带的L2,1范数和沿条带方向的梯度的L0范数实现特征约束;对于干净图像,其边缘连续性采用L1范数进行约束。考虑该约束模型的非凸性,采用基于近端交替方向乘子(PADMM)的算法进行快速求解。模拟数据和实际数据的实验结果表明,所提算法有效去除了图像中的条带噪声,同时高效保留原图像的细节信息,充分验证了算法的可行性和可靠性。To eliminate the stripes of Remote Sensing Image(RSI)which is easy to cause blur or loss of detail the direction and structure characteristics of clean image and stripes are analyzed and studied a stripe separation method for RSI is proposed.For the stripes the sparse structure feature is characterized by the group sparsity and smoothness of the stripes the characteristic constraints are realized by combining the L21 norm of stripe and L0 norm of the gradient along the direction of stripe.For clean image the edge continuity is constrained by L1 norm.Considering the non-convexity of the constraint model the algorithm based on Proximal Alternating Direction Method of Multipliers(PADMM)is used to solve the constraint model fast.Experimental results of simulation data and actual data show that the proposed algorithm can effectively remove the stripes and preserve the details of the original image efficiently which fully verifies the feasibility and reliability of the algorithm.

关 键 词：条带噪声 非凸优化 组稀疏 方向特征 结构特征 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]