SAR image despeckling using nonsubsampled Contourlet transform
- 信息技术－已发表论文 
提出了一种基于无下采样Contourlet变换(NSCT)的合成孔径雷达(SAR)图像去噪方法。首次在理论上证实了SAR图像取对数后无下采样Contourlet系数服从广义高斯分布,从而提出采用贝叶斯阈值方法估计不含噪声的无下采样Contourlet系数,达到去除噪声的目的。仿真和实际实验结果表明,该方法在噪声平滑、边缘和纹理保护等方面优于其他方法。由于无下采样Contourlet变换不进行下采样,该方法能够避免其他进行严格下采样的变换去噪时所引入的人工痕迹。A nonsubsampled Contourlet transform (NSCT) based despeckling method for synthetic aperture radar (SAR) images is presented. It is derived in theory for the first time that the coefficients of log-transformed SAR images decomposed by NSCT obey general Gaussian distribution, so Bayesian shrinkage factor is adopted to estimate noise-free NSCT coefficients. Simulation and experiments demonstrate that the visual quality of the results is superior to other despeckling methods in terms of both background smoothing, preservation of edge sharpness and texture. The absence of decimation in Contourlet decomposition avoids artificial impairments often introduced by other critically subsampled transform methods.