摘要
针对可见光图像弱小目标检测中的背景抑制和去噪问题,提出了奇异值分解(Singular Value Decomposition,SVD)带通滤波新方法.首先分析了图像奇异值与目标、噪声和图像背景的关系,结果表明奇异值的高序部分更多地反映图像噪声,中序部分更多地反映目标性质,而低序部分更多地反映图像背景.以此为依据提出了SVD-Ⅰ型和SVD-Ⅱ型两种带通滤波器,并给出了奇异值曲线转折点法和门限准则法两种滤波器参数确定方法.实验表明SVD带通滤波能有效抑制图像背景,去除噪声,进而提高弱小目标的信噪比.
A new singular value decomposition (SVD) band-pass-filter technology is presented for background suppression and denoising in small targets detecting of visible images. Firstly, the relation between image singular value and targets, image noise and image background is analyzed. And results show that the high order part of image singular value obtains more information of image noise, the middle order part obtains more information of targets and the low order part obtains more information of image background.Based on this fact,two SVD band-pass-fdters named SVD-Ⅰ and SVD-Ⅱare proposed, And two fdter parameters estimation methods are given, including singular value curve turning-point method and threshold criterion method. Experiments show that the new SVD band-pass-fdter can suppress image background and denoising effectively,and improve the signal to noise (SNR) of small targets.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2008年第1期111-116,共6页
Acta Electronica Sinica
基金
国家863高技术研究发展计划(No.2004AA731270)
关键词
背景抑制
图像去噪
奇异值分解(SVD)带通滤波
奇异值曲线转折点法
门限准则法
background suppression
image denoising
singular value decomposition (SVD) band-pass-falter
singular valuecurve turning-point method
threshold criterion method