摘要
为了在抑制噪声的同时更好地保持PET重建图像中的细节结构,提出了一种基于非局部方向性核先验(NSKP)的Bayesian重建算法.为了充分利用图像中的全局信息,该算法在二阶核回归过程中估计出图像梯度,计算出相应的方向性矩阵,并根据非局部均值权值矩阵和方向性矩阵的卷积,计算先验项的权值.在重建中,该算法在高阶核回归过程中同时更新图像的梯度和先验信息,而不是单独计算图像梯度.另外,高阶核回归方法运用多自由度的参数估计提高了重建的精确度.研究结果表明,该算法通过计算引入局部结构信息的全局先验权重,更好地抑制了噪声和过平滑,保持了重建图像中细节区域的结构性和背景区域的一致性.对体模数据的模拟实验结果从视觉和数值角度验证了该算法在PET图像重建中的有效性.
To preserve detail structures and suppress noise for position emission tomography（PET） images,a novel Bayesian reconstruction algorithm based on nonlocal steering kernel prior（NSKP） is proposed.To utilize the global information of the image,the algorithm estimates the image gradient information and calculates the directional matrix in the process of two order kernel regression.Then,the weights for the prior term are calculated from the convolution between the nonlocal means weighting matrix and the directional matrix.During the reconstruction,instead of calculating the gradient of the update image,the NSKP approach estimates the gradient and obtains the prior simultaneously.Furthermore,the more degrees of freedom are used in the high order kernel regression to improve the estimation accuracy.The results show that the reconstruction algorithm uses the global prior with local structure information to overcome the noise and over-smoothness,which preserves the structures in detail regions and the consistency in background regions.The simulation results of phantom data prove the effectiveness of the proposed algorithm in points of visional evaluation and numerical evaluation.
出处
《东南大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2010年第5期937-942,共6页
Journal of Southeast University：Natural Science Edition
基金
国家重点基础研究发展计划(973计划)资助项目(2010CB732503)
江苏省科研创新学者攀登计划资助项目(BK2009012)
国家自然科学基金资助项目(8100636)
关键词
Bayesian-MAP
非局部方向性核先验
方向性矩阵
高阶核回归
结构自适应重建
Bayesian-MAP（maximum a posterior）
nonlocal steering kernel prior（NSKP）
directional matrix
high order kernel regression
structure adaptive reconstruction
