摘要
针对工业断层CT进行三维结构成像的应用需求,提出了由多角度的平行投影数据重建三维CT图像的TV-ART迭代算法的新实现方法,其中将Chambolle方法推广至三维情形并用于求解CT图像全变差(TV)最小。使用该方法进行TV求解的重建图像的质量优于基于最速下降法或共轭梯度法的TV-ART迭代算法。此外,该方法具有高度并行性,适合在GPU,FPGA等高速并行计算硬件上实现,从而可以大幅提高图像重建速度。在扫描时间相同的情况下,该方法重建的三维CT图像质量优于已有方法,特别是显著提高了CT图像的轴向分辨率。
In order to reconstruct 3D CT image from a set of parallel fan-beam projection data acquired at multiangles by industrial computed tomography（ICT） with line-array detector, a novel method for realizing the TV- ART algorithm was proposed. In this method, we adapted the Chambolle＇s method and extended it to 3D condition for solving the Total Variation Minimurh{ TVM） of the 3D CT image, resulting in better image quality than the TV- ART using the steepest descent method or the conjugate gradient method. Meanwhile, this method was suited to implement in GPU or FPGA which was highly parallel so as to increase the calculate speed of reconstruction, The quality of 3D CT image reconstructed by our method was superior to existing methods, especial in the axial resolution.
出处
《无损检测》
2012年第7期11-16,共6页
Nondestructive Testing
基金
国家自然科学基金资助项目(60971131)
北京市自然科学基金资助项目(2011重点B类)
北京市人才强教计划资助项目
