节点文献

基于非局部均值的SAR图像相干斑抑制算法研究

SAR Image Despeckling Based on Nonlocal Means Filter

【作者】 库永霞

【导师】 王卫星; 芦碧波;

【作者基本信息】 河南理工大学 , 计算机应用技术, 2010, 硕士

【摘要】 合成孔径雷达(Synthetic Aperture Radar,SAR)是一种在微波波段工作的相干成像雷达系统,但其固有的成像方式导致图像有斑点噪声出现,为目标识别和特征提取造成了困难。因此,去除相干斑噪声成了SAR图像处理和图像分析之前的关键一步。目前普遍认为,SAR图像中的相干斑为乘性噪声。为去除此类噪声,本文首先利用对数变换将其转化为加性噪声,然后将双边滤波和非局部算法与小波方法结合进行去噪处理,最后再进行指数变换得到去斑后的图像。具体来讲,本文主要包括以下三个方面的内容。(1)针对现有的SAR图像相干斑噪声抑制算法的不足,我们把多分辨率双边滤波算法引入到了相干斑噪声抑制中来。首先对对数转换后的含有加性噪声的图像进行小波分解,这时图像被分解成高频和低频两部分,然后分别在这两部分采用小波阈值和双边滤波的方法进行处理。(2)接着本文引入了基于非局部均值的SAR图像去斑算法,这种算法的核心是对对数转换后的SAR图像采用非局部均值的方法做处理。(3)受到多分辨率双边滤波算法的启发,提出了多分辨率非局部均值方法,并将其应用到了SAR图像去斑中。仿真实验和数据分析表明,上述三种方法在SAR图像相干斑噪声抑制和图像的边缘保持方面具有一定的优势。

【Abstract】 Synthetic Aperture Radar (SAR) is a kind of microwave band coherent radar system. But its inherent imaging mode makes the SAR image inevitably produce a coherent speckle noise (speckle), which seriously influences the quality of the image. It brings difficulties to Target recognition and feature extraction. Therefore, the reduction of speckle became the key step in processing SAR image before the image processing and image analysis. It is widely accepted that speckle of SAR image is multiplicative noise. For removing such noise, firstly, the image of the multiplicative noise is transformed into additive noise. This paper attempts to focus on the use of multiresolution Non-Local means to the additive noise of image. Then the exponentiation is used to the processed image, and can gained suppress speckle of image. Specifically, this paper contains three aspects as follows:Firstly, according to existing SAR image despeckling of disadvantage, speckle denoising with multiresolution bilateral filtering is introduced. First, we use wavelet decomposition to the logarithmic conversion image, when the image is decomposed into high and low frequencies. Then the two parts can make use of seting thresholds and bilateral filtering methods.Secondly, based on Non-Local means of SAR images is introduced.we use the non-local mean filtering method for the image of additive noise.Thirdly, inspired by multiresolution bilateral filtering, a multiresolution Non-Local means is proposed and applied to the despeckling of the SAR image.The data analysis and the simulation experiments of the paper in each section show: the speckle denoising algorithms of SAR image and retaining the edge have certain advantages.

节点文献中: 

本文链接的文献网络图示:

本文的引文网络