用于超声斑点噪声滤波的各向异性扩散新模型  

A New Speckle Reducing Anisotropic Diffusion for Ultrasonic Speckle

作　　者：李灿飞[1] 王耀南[1] 肖昌炎[1] 卢笑[1] 

机构地区：[1]湖南大学电气与信息工程学院,长沙410082

出　　处：《自动化学报》2012年第3期412-419,共8页Acta Automatica Sinica

基　　金：国家高技术研究发展计划(863计划)(2007AA04Z244;2008AA04Z214);国家自然科学基金(61172160;60835004;60872130);湖南省自然科学基金(09J13118);中央高校研究基金(53110704032)资助~~

摘　　要：由于扩散系数的缺点,原斑点噪声各向异性扩散模型(Speckle red ucing anisotropic diffusion,SRAD)有产生板块效应、模糊弱边界与细节等缺点.本文通过改进扩散系数,提出一种新的斑点噪声各项异性扩散模型(New speckle reducing anisotropic diffuse,NSRAD),NSRAD中采用一个S型函数作为扩散系数:在同质区域中,采用各向同性扩散,避免了板块效应;在结构性区域中,扩散速度变化敏感,同时以更快趋向于0的速度扩散,因此,提高了该区域的分辨率,达到增强细节和弱边界以及保留边界的锐利性的目的.仿真图像的定量分析证明新方法不仅比原SRAD除噪更有效,而且提高了除噪后图像与原图像的结构相似性,同时具有更小形变.真实图像的试验结果也证明新方法在有效除噪的同时消除了黑板刷效应,增强了边界以及细节.Although speckle reducing anisotropic diffusion （SRAD） has been greatly explored in ultrasonic imaging, it generates block effects in homogeneous regions and blurs edges due to the intrinsic diffusion. To address these defects, we propose a new speckle reducing anisotropic diffuse （NSRAD） in which a new Sigmoid function as the diffusion coefficient is suggested. Due to the specific properties of such novel diffusion coefficient at various region, NSRAD is able to diffuse： 1） in a nearly constant speed within homogeneous regions, thus to avoid the block effect; 2） in a sensitively varying speed within the transition regions, at the same time, in a speed approaching to zero more quickly than that in SRAD concerning the structure regions, thus helps to improve the resolution and enhance the weak edge while sharpening the strong edges at the edges. The performances of the proposed model are verified in experiments with synthetic and clinical images. Especially, a quantitative comparison between NSRAD and SRAD is made upon simulation data, where we demonstrate that NSRAD suppresses speckle more efficiently in homogenous regions hereby has less distortion, while in inhomogeneous regions it enhances details and sharpens the edges more effectively than SRAD.

关 键 词：各向异性平滑 斑点滤波 超声图像 局部统计 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]