参考图像法CT迭代重建  

Reference Image Method CT Iterative Reconstruction

作　　者：莫仕灯 王朕 华斌[1] MO Shi-deng;WANG Zhen;HUA Bin(Department of Information Science and Technology,Tianjin University of Finance and Economics,Tianjin 300222,China)

机构地区：[1]天津财经大学理工学院,天津300222

出　　处：《核电子学与探测技术》2019年第6期774-780,共7页Nuclear Electronics & Detection Technology

基　　金：天津市自然科学基金(16JCYBJC42000)资助。

摘　　要：介绍了应用参考图像的CT迭代重建算法。使用拉格朗日对偶方法将约束最小化问题转移到相应的非约束最大化对偶问题,将重建图像与参考图像的L_P范数组成的距离函数作为图像重建的代价函数,使用梯度上升法多次迭代以优化代价函数。同时,提出了四种参考图像的选取方案。实验结果表明:在较少的迭代次数下,有参考图像的重建图像质量好于无参考图像的重建图像质量。其中,若选取FBP重建图像作为参考图像则容易获取且不需要知道扫描对象的预期结构;若选取拼接图像作为参考图像则重建图像的质量显著提升,条纹型的失真伪影明显被抑制,而真实的图像细节依旧可以保留。The CT iterative reconstruction algorithm for applying reference images is introduced.Using the Lagrangian dual method to transfer the constraint minimization problem to the corresponding unconstrained maximization dual problem,the distance function composed of the reconstructed image and the LP norm of the reference image is used as the cost function of the image reconstruction,multiple iterations are performed using the gradient ascent method to optimize the cost function.At the same time,four selection schemes of reference images are proposed.The experimental results show:at fewer iterations,the reconstructed image quality of the reference image is better than the reconstructed image quality without the reference image.Wherein,if the FBP reconstructed image is selected as the reference image,it is easy to obtain and doesn’t need to know the expected structure of the scanned object;if the stitched image is selected as the reference image,the quality of the reconstructed image is significantly improved,the stripe-shaped distortion artifact is obviously suppressed,and the real image details can still be retained.

关 键 词：参考图像 迭代重建算法 低剂量 LP范数 CT 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]