摘要
提出一种基于小波多尺度相关的目标检测方法。该方法通过二进小波对图像进行多尺度分析;利用边缘和噪声具有不同的Lipschitz指数造成它们的小波变换模在不同尺度下的不同传播特性,根据小波变换尺度相关性计算相邻大尺度模相关量用来增强信号,抑制噪声,提取粗略边缘图;并结合小尺度检测,具有定位精确的特点,将粗略边缘图与小尺度模图相与,寻找模极大值点,得到最终的检测结果。实验结果表明,该方法较传统的小波多尺度边缘检测方法能更加有效地进行红外小目标检测。
Based on wavelet scale correlations, a new algorithm of infrared small target detection is presented. The discrete dyadic wavelet transform was employed to produce the multiscale representation of image, correlations of big scales＂ modulus are calculated based on the difference characterization of edges and noise by Lipschitz exponents and a coarse edge map was then scavenged, logical-AND operation was performed between the coarse edge map and the modulus map of small scale and the maxima of the wavelet transform modulus were checked out, and finally, infrared small target was detected. Experiments turn out that the algorithm works more efficiently in infrared small target detection than the original multi-scale edge detection one.
出处
《半导体光电》
EI
CAS
CSCD
北大核心
2007年第4期592-595,共4页
Semiconductor Optoelectronics
基金
国家"863"计划资助项目
