基于NSCT距离像的星空背景小目标检测技术  

A Method of Small Target Detection in the Starry Sky Background based on NSCT Distance Map

作　　者：何宇 杨馥溢 杨璐[2] 

机构地区：[1]成都信息工程大学信息中心,四川成都610225 [2]成都信息工程大学计算机学院,四川成都610225

出　　处：《成都信息工程大学学报》2017年第3期264-268,共5页Journal of Chengdu University of Information Technology

基　　金：四川省教育厅基金资助项目(14ZA0176)

摘　　要：点状空间目标和长条状背景星存在几何特征的不同,在频域中表现为:长条状背景星主要分布某方向子带上,而点状目标在所有方向子带上的分布近似相同。非下采样Contourlet(NSCT)变换可以对图像进行灵活的多尺度、多方向和平移不变性的分解,能很好地捕捉长条状轮廓,文中提出的基于NSCT距离像的目标检测技术中将某尺度上各方向分量间的比值作为特征向量,通过计算它与中心向量的距离得到距离像,从距离像的统计直方图中可以自适应地选择分割点状目标的门限,实验证明该方法能有效地检测图像中的点状小目标。There are some differences of geometry features between the points-like small target and the strip-like background stars. In the view of frequency analysis,the background stars will have a main direction band in the frequency domain,because it has a Strip-shaped look,while the points-like small target can be treated as pulse signal,so its energy will distribute all the band of frequency domain. The Nonsubsampling Contourlet Transform（ NSCT） hasflexible multiscale,multidirection,and shift-invariant image decomposition characteristics,so it can efficiently capture the differences of the geometry features. This paper proposes a detection method based on the NSCT Distance Map： defined the ratios between each direction component as a feature vector,and then get a distance image by calculating the distance between the feature vector and the center vector,after that,can adaptively get the threshold that used to segmented the small target from the image from the grayscale histogram. Experiments show that the method can efficiently detect the small target from the image.

关 键 词：星空背景 非下采样CONTOURLET变换 方向子带 距离像 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]