一种基于小波域隐马尔可夫模型的SAR相干斑抑制算法  

An Algorithm Based on Wavelet-Domain Hidden Markov Models for SAR Speckle Reduction

作　　者：武昕伟[1] 朱兆达[1] 张弓[1] 郭春生[1] 

机构地区：[1]南京航空航天大学信息科学与技术学院401室,南京210016

出　　处：《中国图象图形学报（A辑）》2004年第4期385-390,共6页Journal of Image and Graphics

基　　金：国家自然科学基金项目(60082002);江苏省自然科学基金项目(BK2001047)

摘　　要：相干斑噪声是 SAR图像的固有特点。对相干斑抑制的要求是在平滑噪声的同时 ,尽量保持原始图像的结构信息。现有的许多相干斑抑制方法各有优点和不足 ,没有普遍的适用性。基于图像在小波域的隐马尔可夫模型(HMMs)结构 ,结合 SAR图像中相干斑噪声的统计特性 ,本文提出了一种新的小波域相干斑抑制方法。仿真及实测数据处理结果表明 ,该方法在有效抑制相干斑的同时 ,更好地保持了边缘结构。与小波域软阈值去噪方法和 L ee滤波器相比较 ,该方法在噪声平滑及边缘保持上都取得了较大的改进 。Speckle noise is an intrinsic property of Synthetic Aperture Radar (SAR) imagery. The demand for speckle reduction of SAR images is to smooth the speckle noise while preserving the structure information of the original images. Existing speckle suppression methods possess respective merits and drawbacks, without universal adaptability. Integrating the statistical characteristic of speckle noise in SAR images with wavelet domain hidden Markov models (HMMs) structure of images, we propose a new wavelet domain speckle reduction method. Simulation and experimental results using real data show that the proposed method is able to effectively suppress speckle noise and to better retain edge structure. Compared with wavelet domain soft thresholding denoising algorithm and Lee multiplicative speckle filter, the wavelet domain HMMs method offers significant improvements on smoothing speckle and preserving edge. In addition, the proposed method also gets a better visual effect.

关 键 词：合成孔径雷达 相干斑 小波变换 隐马尔可夫树模型 SAR 后向散射系数 

分 类 号：TN958[电子电信—信号与信息处理]