基于贝叶斯模型的shearlet域SAR图像去噪方法

摘要

A shearlet domain synthetic aperture radar(SAR) image denoising algorithm based on Bayesian model is presented, through the characteristic analysis of the SAR image noise.Firstly, the SAR image in the shearlet domain is represented sparsely to obtain the distribution of the sparse coefficient.Secondly, the signal and noise detection modeling is carried out by using the Bayesian model to solve the problem of the optimal threshold.Then, the SAR image noise is smoothed by using the adaptive weighting algorithm, according to different characteristics of the correlation of the sparse coefficient in different directions.Finally, conducting the inverse shearlet transform by using the high and the low frequency sub images of the noise reduction to obtain the SAR reconstruction image.The experimental results show that the proposed algorithm can suppress speckle, as well as can restrain the image edge information better by means of the experiment in MSTAR database.%通过对合成孔径雷达(synthetic aperture radar,SAR)图像相干斑噪声的特点分析,提出一种基于贝叶斯模型的shearlet域SAR图像去噪方法.首先将变换后的SAR图像在shearlet域进行稀疏表示,得到稀疏系数的分布;其次利用贝叶斯模型进行信号和噪声检测的建模,得到最佳的阈值;然后根据稀疏系数在不同方向上相关性不同的特点,利用自适应加权收缩算法对SAR图像噪声进行平滑处理;最后利用降噪后的高频子图像和低频子图像进行逆shearlet变换,得到SAR重构图像.通过在MSTAR数据库上的实验表明,该算法在滤除相干斑噪声的效果上比其他方法更好,并且不会损失图像的边缘特性.