摘要
针对合成孔径雷达(SAR)图像固有的相干斑噪声,提出了基于小波域多尺度隐马尔可夫模型(HMM)的去噪方法.该方法首先分析了小波域系数的统计特性,利用B样条小波基所生成滤波器的线性相位性对图像系数进行了统计建模,通过将1D信号的处理技术应用到2D信号,实现了图像系数建模更为准确、参数训练速度更快、斑点噪声抑制更加有效的目的.与常用的几种滤波算法相比,实验结果也表明该方法在平滑噪声和保持有用信号细节两方面均显示出了较好的效果.
An efficient method based on multiresolution hidden Markov model of wavelet-field to suppress the speckle noise in synthetic aperture radar (SAR) images is presented. It begins with an analysis of the coefficient statistic characteristics of wavelet-field. The image coefficient is modeled by using linear phase of statistic B-spline base filter. The two-dimension signal processing method is applied instead of one-dimension. Thus the more accurate modeling of image coefficients, the faster speed of training parameter and the more effective suppressing of speckle noise are realized. Compared to several conventional denoising methods, the simulative experimental results show that it has obvious merits in noise smoothing and available signal detail keeping as well.
出处
《武汉大学学报(工学版)》
CAS
CSCD
北大核心
2005年第3期126-130,共5页
Engineering Journal of Wuhan University