摘要
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。
In the paper, additive model was introduced to SAR image, and Non-Subsampled Contourlet Transformation was used. According to the correlation among its coefficient, neighborhood coefficients and the father coefficients, it proposed a classification standard where the coefficients were divided into the important and unimportance coefficients. Moreover, the improved Donoho＇ s threshold was used to process the important coefficients, and estimated the original NSCT coefficients from noisy coefficients. Thus the speckle .was removed of SAR image. Lastly, experimental results for speckle reduction of real SAR images showed that the algorithm would outperform the other current despeckling algorithms both in the despeekle effect and the detail reserve of image.
出处
《测绘科学》
CSCD
北大核心
2013年第5期77-79,83,共4页
Science of Surveying and Mapping
基金
安徽省教育厅重点科研项目(KJ2010A282)
安徽省自然科学基金资助项目(11040606M06)
