SAR image despeckling using nonsubsampled Contourlet transform

摘要

中文摘要: 提出了一种基于无下采样 Cont ourlet变换 (NSCT)的合成孔径雷达 ( S AR)图像去噪方法。首次在理论上证实了 S AR图像取对数后无下采样 Cont ourlet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 Cont ourlet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 Cont ourlet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。英文摘要：Abstract: A nonsubsamp led Cont ourlet transform (NSCT) based des peckling method for synthetic aperture radar( SAR) i mages is p resented . It is derived in theory for the first ti me that the coefficients of l ogtransfor med SAR images decomposed by NSCT obey general Gaussian distributi on, so Bayesian shrinkage fact or is adop ted t o esti matenoise free NSCT coefficients . Simulati on and experi ments demonstrate that the visual quality of the results is superi orto other des peckling methods in terms of both backgr ound s moothing, preservati on of edge sharpness and texture .The absence of deci mati on in Cont ourlet decompositi on avoids artificial i mpair ments often intr oduced by other critically subsamp led transfor m methods .