论文部分内容阅读
本文对SAR图像在引进加性模型的基础上,采用非下采样Contourlet变换,根据其系数、邻域系数及父系数3者之间的相关性,给出一个分类准则,把系数分为2类:重要系数和非重要系数,然后采用改进的Donoho阈值处理重要系数,估计出不含噪声的非下采样Contourlet变换系数,从而得到抑制了相干斑的SAR图像。对真实SAR图像进行相干斑噪声抑制实验,结果显示本文方法在抑斑效果和图像的细节保留上均优于目前的许多方法。
Based on the introduction of additive model and the nonsubsampled Contourlet transform, this paper presents a classification criterion based on the correlation between its coefficients, neighborhood coefficients and the paternity coefficients, and divides the coefficients into two categories : Significant coefficients and non-significant coefficients, and then use the improved Donoho threshold to process the important coefficients, and estimate the non-subsampled Contourlet transform coefficients without noise to obtain SAR images with speckle suppression. Experiments on speckle noise suppression of real SAR images show that the proposed method is superior to many current methods in speckle reduction and image detail preservation.