摘要
由于遥感图像数据量非常大,在使用投影寻踪进行图像的无监督分割时存在计算量大的缺点.本文证明了服从正态分布的独立随机变量序列经小波变换后,其低频系数与高频系数仍服从正态分布.并利用该性质将低频系数用于寻找投影寻踪方向,由于低频系数经过了二进抽样,从而减少了寻找投影方向的计算量.从实验结果可以看出该方法是快速有效的,在对海量数据的遥感图像实时分割中具有较强的实用价值.
Because the remote sensing image often has large number of data,the large computation amount should been used in using projection pursuit to segment this kind of image.If independent random sequence has normal distribution,then the low frequency coefficients and high frequency coefficients getting by the discrete wavelets transformation of this sequence are still follow norm of distribution.And the method how to find out the direction of projection quickly by the low frequency coefficients is proved.This method needs less time to find out the direction,because the low frequency coefficients have fewer data.It is valuable to process the remote sensing image in real time.
出处
《湖南理工学院学报（自然科学版）》
CAS
2010年第4期18-21,共4页
Journal of Hunan Institute of Science and Technology(Natural Sciences)
关键词
投影寻踪
遥感图像
小波变换
图像分割
projection pursuit
remote sensing image
wavelets transformation
image segmentation
