摘要
针对传统离散模型对Radon变换的离散化存在误差这一缺点,引入基于理想区域的离散重建点模型,以减小离散化模型带来的误差,使重建过程更加准确.该模型基于物理过程对线积分进行离散化,以重建点理想区域来代替传统意义上的像素格,用基函数来刻画局部理想区域的密度分布;然后通过对基函数加窗来抑制频谱混叠,从而得到连续平滑的重建图像;最后用EM算法来完成新模型的迭代重构.实验结果表明:在相同迭代次数下,新模型下的重建误差较之传统模型有所减小,窗函数对误差也起到了抑制作用,验证了新模型下EM迭代算法的有效性.
The traditional discrete model existed errors in discretization of continuous Radon transform. Based on reconstruction point, a new model was introduced, which reduced errors caused by the discrete model and made the reconstruction more accurate. The discretization of line integral was based on physical process, and the pixel grid was replaced by ideal area, whose density distribution was described by basis function, then the basis function was windowed in order to suppress frequency aliasing, thus the reconstructed image would become smooth. Finally, the new model was reconstructed by EM iterative algorithm. The experimental results illustrate that the error of the new model is reduced compare with the traditional model under the same number of iterations, and windowed basis function also plays a role in restraining the errors, which verifies that the EM iterative algorithm on new discrete model is effective and feasible.
出处
《中北大学学报（自然科学版）》
CAS
北大核心
2014年第2期209-217,共9页
Journal of North University of China(Natural Science Edition)
基金
国家自然科学基金资助项目(61071192
61271357
61171178
60972115
61271425)
山西省自然科学基金资助项目(2009011020-2)
山西省国际合作项目(2013081035)
山西省高等学校优秀青年学术带头人支持计划资助课题
