基于权值核范数最小化的红外背景杂波抑制  

Infrared Background Clutter Suppression Based on Weighted Nuclear Norm Minimization

作　　者：司马端 安玮[1] 王普[1] 龙云利[1] 

机构地区：[1]国防科技大学电子科学与工程学院,湖南长沙410073

出　　处：《数字技术与应用》2015年第12期72-74,共3页Digital Technology & Application

摘　　要：针对复杂背景下的空间背景抑制问题,提出基于权值核范数最小化理论的单帧红外图像空域背景杂波抑制算法。首先利用红外图像的非局部相似性质,将具有相似性的图像块向量化并聚合成低秩矩阵,然后对所得的低秩矩阵的奇异值赋予不同的权值,最后利用加权核范数最小化将背景抑制问题转化为优化问题进行求解。预测的背景在满足对观测图像数据依赖的同时,能够保留背景图像的边缘信息,有效降低复杂背景灰度值起伏较大处的虚警。通过仿真实验验证了算法的可行性和有效性,表明其背景抑制性能较传统算法有较大提高。An algorithm of spatial background clutter suppression of infrared image based on weighted nuclear norm minimization theory is proposed for small target detection in complicated background. Firstly,by exploiting the image nonlocal self-similarity,the nonlocal self-similar image patch vectors are stacked into a low rank matrix.Then, the singular values oflow rank matrix are assigned different weights.Finally, by using weighted nuclear norm minimization converting it to an optimization problem. While the data fidelity to observed image holds, the background estimate can still maintain edges, effectively reducing false-aiarm where gray level varies dramatically in complicated background. Feasibility and effectiveness of the algorithm is verified by simulation experiments. Analysis of the result shows that background suppression performance is remarkably improved compared to traditional method.

关 键 词：非局部相似 加权核范数 背景抑制 红外小目标 

分 类 号：TN97[电子电信—信号与信息处理]