海洋合成孔径雷达图像斑点噪声的滤除  

Speckle noise filtering for sea Synthetic Aperture Radar images

作　　者：蒋永馨[1,2] 王孝通[2,3] 徐晓刚[2,4] 黄华[2,3] 

机构地区：[1]海军大连舰艇学院基础部,大连116018 [2]海军大连舰艇学院光电技术研究所,大连116018 [3]海军大连舰艇学院航海系,大连116018 [4]海军大连舰艇学院装备系统与自动化系,大连116018

出　　处：《高技术通讯》2009年第6期586-590,共5页Chinese High Technology Letters

基　　金：国家自然科学基金(6047314);辽宁省自然科学基金(20082176);浙江大学CAD&CG国家重点实验室开放基金资助项目

摘　　要：针对海洋合成孔径雷达(SAR)图像结构信息不明显、水下目标识别困难这一问题,提出了基于经验模式分解(EMD)的算法和H(o|¨)lder指数调整相融合的框架,该框架可以有效地滤除海洋SAR图像的斑点噪声并增强其结构信息,使得人眼可以分辨其特征信息。该融合框架利用EMD将海洋SAR图像分解成不同频率成分的分量,不同层次的分量根据其结构信息和噪声的特征用不同的H(o|¨)lder指数来调整,H(o|¨)lder指数的大小随着分量层数的增加而减小,即在不同尺度下分别抑制斑点噪声,从而恢复其中所包含的结构信息。试验结果表明,利用该框架可以有效抑制SAR图像中的斑点噪声和增强与水下目标相关的结构信息,使人眼可以分辨海洋SAR图像的特征结构。To solve the problem that underwater objects are difficult to recognize with the method of Synthetic Aperture Radar （SAR） due to the insufficiency of the structure character information of sea SAR images, the paper proposes a noise elim- ination algorithm based on combining the experiment mode decomposition （EMD） algorithm with the Holder Exponent ad- justment. The algorithm can effectively suppress the speckle noise and enhance the structure character information, which finally become visible to the naked eye. Using the different frequency character of each level, the content of each level noise and non-noise are analyzed and the Holder exponent of each level is adjusted. And the speckle noise of simple SAR images in different scale is filtered. At last, the new denoised SAR images can be obtained. The experimental result in- dicates that the speckle noise is substantially decreased and the structure character and texture of the result is clearer. And human eyes can distinguish the structure character of sea SAR images.

关 键 词：合成孔径雷达(SAR) 图像去噪 斑点噪声 Holder指数 

分 类 号：TN957.52[电子电信—信号与信息处理] P732.6[电子电信—信息与通信工程]