【摘要】 针对合成孔径雷达图像斑点噪声去除问题 ,提出了一种滤波算法。该算法根据 SAR图像小波系数分布的特点 ,采用高斯混合模型对其进行精确拟合 ,并用贝叶斯估计来恢复原图。为了有效地克服对数图像均值不为零对小波滤波的影响 ,在多尺度小波变换前先对取对数后的图像作了归一化处理。仿真实验结果表明 ,该方法克服了常用的小波阈值去噪只有一个阈值的缺点 ,可在去除噪声的同时有效保留图像的边缘细节更多还原

【Abstract】 A new algorithm for suppressing speckle in synthetic aperture radar (SAR) images is presented. According to the statistical characteristics of wavelet coefficients, Gauss mixture model is used to demonstrate the statistics of those coefficients. Using Bayesian estimator is to get the actual coefficients. Biased mean of the log-transformed image will affect the algorithm result; it is corrected before multi-scale wavelet transformation. Simulation results show that the method performs better than the common threshold de-noising method. So it can remove the noise while sustaining the edge properties of natural SAR images.更多还原