基于方向波域混合高斯模型的SAR图像去噪

SAR IMAGE DENOISING BASED ON GAUSSIAN MIXTURE MODEL IN DIRECTIONLET FIELD

为了解决传统去噪方法处理SAR(Synthetic Aperture Radar)图像存在的失真、图像边缘模糊等问题,提出一种新的去噪算法。首先对经过对数变换后的SAR图像进行方向小波变换,并对无噪图像的方向小波系数建立混合高斯模型,再用矩估计法对其进行参数估计,最后通过贝叶斯滤波器去除噪声。实验结果表明,该算法可以得到较好的去噪效果,并且有效地解决传统去噪算法在图像失真、边缘模糊等问题存在的不足。

To solve the problems of distortion and edge burring the traditional SAR （ synthetic aperture radar） images denoising methods have, a new denoising method is proposed. First, the SAR images that have been processed by logarithmic transformation are processed with directional wavelet transform. Then a Gaussian mixture model is built for the directionlet wavelet coefficients of noiseless images, and the corresponding parameters are estimated with moment estimation. Finally, the noise is removed by the Bayesian filter. Experimental results show that this algorithm can achieve better effect of denoising and can effectively solve the insufficiencies of traditional denoising method in image distortion and edge burring.