基于directionlets变换的SAR图像相干斑噪声抑制算法研究

项目名称： 基于directionlets变换的SAR图像相干斑噪声抑制算法研究

项目编号： No.60872163

项目类型： 面上项目

立项/批准年度： 2009

项目学科： 化学工业

项目作者： 高清维

作者单位： 安徽大学

项目金额： 32万元

中文摘要： Directionlets 变换是一种新的多尺度几何分析方法，能够有效刻画图像的几何结构，充分体现图像数据的各向异性特征，directionlets 变换在数值实现上具有和二维离散小波变换相同的复杂性，计算简单快速；利用directionlets 变换对SAR 图像进行多尺度、多方向分析，形成低频近似图像及多尺度、多方向的directionlets 变换域高频细节图像；通过对directionlets 变换域细节图像信号的研究,选择合适的数理统计模型，利用Bayes 估计准则，从噪声污染的变换系数中估计出"干净"的directionlets 变换系数，再进行directionlets 逆变换重建图像，以达到抑制相干斑噪声的目的，在实现SAR 图像的相干斑去除的同时，能很好的保留图像边缘细节，便于SAR 图像后续的目标检测、识别、分类等应用处理。

中文关键词： directionlets变换；SAR图像；相干斑噪声

英文摘要： Directionlets transform is a new multi-scale geometrical analysis method, which can capture the intrinsic geometrical structure in natural images, and exhibit anisotropic characteristic of image data. The numerical computation of this transform is fast and simplicity and has the same computational complexity comparing with two dimensional discrete wavelet transform. SAR image contaminated by coherent speckle is decomposed by multi-scale and multi-resolution directionlets transform, approximation and directionlets details appear; suitable statistical model is selected for these directionlets details through studying the multi-scale, multi-resolution details, and the clear directionlets details are also estimated by Bayes estimation principle, then inverse directionlets transform is implemented, the coherent speckle in SAR image is suppressed. This method suppresses the speckle noise effectively, and preserves many target characteristics of original images. The despeckled SAR image is advantageous to the target detection, recognition and classification.

英文关键词： directionlets transform; SAR image; coherent speckle noise