摘要
在分析毫米波辐射图像特性的基础上,对基于单帧的二维最小均方误差(Two-dimensional LMS)背景预测算法进行改进,充分利用连续帧图像中小目标的运动信息和背景的缓变特性,设计出基于连续帧的TDLMS滤波器。针对TDLMS算法对步长μ选取困难的问题,本文采用变步长TDLMS自适应算法。实验证明,该滤波算法能在两帧图像时达到最优的背景预测效果,有效滤除杂波,提高图像信杂比。
According to the characteristics of passive millimeter-wave images,the Two-dimensional Least Mean Square（TDLMS）algorithm is improved,and a TDLMS filter based on consecutive frames is designed,which makes full use of motion information of small target and slow changes of the background in consecutive frames.For the difficult of selecting step-size μ in TDLMS algorithm,adaptive variable step-size algorithm is adopted.Experiments show that this algorithm can achieve the best prediction results of background in two images,removes clutter and improves the image SCR（signal to clutter rate） effectively.
出处
《微波学报》
CSCD
北大核心
2011年第3期1-4,共4页
Journal of Microwaves
基金
国家自然科学基金(40906089)
中国航天科技集团公司航天科技创新基金重点项目(CASC200902)
关键词
无源毫米波成像
背景预测
二维最小均方差
相邻帧
变步长
passive millimeter-wave image
background prediction
TDLMS
consecutive frames
variable step-size
