SAR图像的非下采样Contourlet噪声抑制算法

非下采样Contourlet变换克服了小波变换非一维奇异性最优基的缺点,而且其平移不变性使其边缘保持能力优于Contourlet.提出了一种非下采样Contourlet的合成孔径雷达(SAR,Synthetic Aperture Radar)相干斑去噪模型.首先对SAR图像进行同态变换,将乘性噪声变为加性噪声;再进行非下采样Contourlet变换,包括非下采样金字塔分级和非下采样方向滤波2部分;最后利用阈值的方法将信噪实施分离.实验中针对乘性噪声污染的航拍图像和原始SAR图像,进行了传统的Lee滤波器和小波滤波器去噪,与上述模型去噪进行比较,在去噪性能和边缘保持的主观视觉上都得到了优异的结果.

Nonsubsampled Contourlet speckle reduction algorithm for SAR images

The nonsubsampled Contourlet(NSCT) not only overcomes the disadvantage of wavelet,the nonoptimal basis for one-dimensional singularity,but also improves the edge preservation for the shift-invariance.Therefore,a speckle reduction model based on NSCT was presented.Firstly,original image with multiplicative noise was transformed into with additive noise by means of homomorphic transform.Then,the NSCT was implemented,including two steps which were nonsubsampled pyramids and nonsubsampled directional filter banks orderly.Finally,threshold denoising method was adopted to separate the noise and signal.The simulation experiments were carried out by traditional Lee filter,wavelet filter,Contourlet filter and the above NSCT denoising model.A comparison between them was given out for remote sensing images contaminated by multiplicative noise and synthetic aperture radar(SAR) original image.Experiment results show that the performance of the NSCT is superior not only in speckle reduction but also in edge preservation.