摘要
复杂背景的抑制是红外弱小目标检测技术的一个难题。为解决这个问题,提出了基于奇异值分解的背景抑制算法。从矩阵的角度出发,通过对原图像进行奇异值分解,将包含弱小目标信息的图像矩阵分解到一系列奇异值和奇异值矢量对应的子空间中,然后通过定义的偏差指数所确定的有效的奇异值来重构图像,从而达到背景抑制的目的。与二维最小均方误差算法比较,实验结果显示,该算法对红外弱小目标复杂背景从主观视觉和数值指标都具有良好抑制效果。
Complicated background suppression is a difficult problem for infrared dim and small target detection technique. The algorithm based on singular value decomposition is presented. The algorithm decomposes image which contains dim and small target information to a series of sub-spaces corresponding to singular value and singular value vector. Then, according to bias index, specific effective singular values are extracted to reconstruct target image for background suppression. Compared with conventional Two-dimensional LMS algorithm, through subject inspection and value index, several groups of experimental results demonstrate that the presented algorithm can suppress complicated background in infrared dim and small target image effectively.
出处
《半导体光电》
CAS
CSCD
北大核心
2009年第3期473-476,共4页
Semiconductor Optoelectronics
基金
教育部科学技术研究重点项目(108114)
关键词
红外图像
目标检测
背景抑制
奇异值分解
偏差指数
infrared image
target detection
background suppression
singular value decomposition
bias index