基于细节信息约束的遥感影像条带噪声去除模型  

Remote sensing image stripe noise removal model based on detail information constraints

作　　者：王密[1] 董滕滕 彭涛 项韶 兰穹穹 WANG Mi;DONG Tengteng;PENG Tao;XIANG Shao;LAN Qiongqiong(State Key Laboratory of Information Engineering in Surveying,Mapping and Remote Sensing,Wuhan University,Wuhan 430079,China;China Centre for Resources Satellite Data and Application,Beijing 100094,China)

机构地区：[1]武汉大学测绘遥感信息工程国家重点实验室,湖北武汉430079 [2]中国资源卫星应用中心,北京100094

出　　处：《测绘学报》2024年第9期1799-1816,共18页Acta Geodaetica et Cartographica Sinica

基　　金：国家重点研发计划(2022YFB3902804)。

摘　　要：遥感影像在获取过程中会经常受到条带噪声的污染,降低遥感影像的视觉效果,对影像解译和反演等处理产生不利影响。当前一些主流的基于变分的条带噪声去除方法,虽然可以去除条带噪声,但是往往也会导致影像细节信息的严重丢失。基于上述问题,本文提出了一种基于细节信息约束的遥感影像条带噪声去除模型(DISUTV)。在DISUTV模型中,将所提出的基于双边滤波器与正交子空间投影的细节信息分离算子与单向全变分正则化项、群组稀疏正则化项及单向全变分正则约束项进行了有效结合,并采用交替方向乘子法对其进行求解,用于从条带噪声影像中获取不含有细节信息的高精度条带噪声。利用模拟数据与真实数据对本文方法的条带噪声去除能力、细节信息保持能力及稳健性进行了验证并与现有前沿方法进行了比较。试验结果表明,本文方法在去除条带噪声的同时能更好地保留影像的细节信息,并且呈现出了较好的定性与定量结果。Remote sensing images are often contaminated by stripe noise during the acquisition process,which reduces the visual effect of remote sensing images and has an adverse effect on image interpretation and inversion.Although some mainstream stripe noise removal methods based on variational methods can remove stripe noise,they often lead to serious loss of image detail information.Based on the above problems,this paper proposes a remote sensing image stripe noise removal model DISUTV based on detail information constraint.In the DISUTV model,the proposed detail information separation operator based on bilateral filter and orthogonal subspace projection is effectively combined with one-way total variation regularization term,group sparsity regularization term and one-way total variation regularization constraint term,and the alternating direction multiplier method is used to solve it,which is used to obtain high-precision stripe noise without detail information from stripe noise images.The stripe noise removal ability,detail information retention ability and robustness of the algorithm are verified using simulated data and real data,and compared with existing cutting-edge methods.Experimental results show that the proposed method can better retain the detail information of the image while removing stripe noise,and presents good qualitative and quantitative results.

关 键 词：条带噪声提取 正交子空间投影 细节信息分离算子 单向全变分 群组稀疏 

分 类 号：P237[天文地球—摄影测量与遥感]