基于MAP-MRF的旋转3维超声心动图断层重建及降噪  

Reconstruction and Restoration for Rotation-scanning Three-dimensional Echocardiographic Images Based on MAP-MRF

作　　者：郭强[1] 许健[1] 杨新[1] 吴兰萍[2] 孙琨[2] 

机构地区：[1]上海交通大学图像处理与模式识别研究所,上海200030 [2]上海第二医科大学附属新华医院上海儿童医学中心,上海200092

出　　处：《中国图象图形学报》2005年第10期1281-1288,共8页Journal of Image and Graphics

基　　金：国家自然科学基金项目(30170264);国家重点基础研究发展规划"973"项目(2003CB716104)

摘　　要：提出了一种基于最大后验概率准则(MAP)-马尔可夫随机场(MRF)的3维超声图像重建算法,在3维重建的过程中可以有效地去除超声图像特有的斑点噪声。首先提出了3维重建的数学模型,它假设所要估计的空间灰度值属于某个函数向量空间,这个向量空间的基函数因支持区域的不同而不同,将Rayle igh分布和G ibbs分布作为先验知识,通过后验概率最大化(MAP)来估计基函数前的系数,在重建算法中,通过将邻点联系强度α设定为随梯度的变化而变化,从而达到了保边界的各向异性滤波特性。采用ICM算法求解这个系数向量。为了提高该算法的效率,又进一步提出了其改进算法,最后的实验结果显示,这种基于MAP-MRF的重建算法可以有效地去除超声图像中的斑点噪声。A new 3D reconstruction method for three-dimensional echocardiographic based on MAP-MRF is proposed in this paper. Not only can this method reconstruct 3D echocardiographic images, but also removes the noise, which may contaminate the image during the acquisition process. First a statistical model of reconstruction of three-dimensional echocardiographic images is established in the paper. It assumes that the function f（x） describing the intensity belongs to finite dimension vector space F with known basis functions. Then with the Rayleigh and Gibbs distributions working as observation model and prior model, MAP-MRF method is adopted to estimate the coefficients of basis function. By setting parameter α, the measurement of the connection strength of neighborhood node, as the function of the local gray level gradient, this statistical model is anisotropic, which avoid border blur during the reconstruction process. Furthermore the ICM algorithm is implemented to compute the results. To enhance the performance of the ICM algorithm, a linearized algorithm is proposed in the paper to simplify the calculation. Finally, the results of comparison among this new method and traditional methods in the experiment show that the new method works well.

关 键 词：最大后验概率准则-马尔可夫随机场 3维超声心动图 3维重建 降噪 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]