基于多项式拟合的C臂投影全局校正法  

Polynomial fitting-based global correction method for C-arm projection images

作　　者：王田苗[1] 刘文勇[1] 胡磊[1] 孙磊[2] 张利军[1] 

机构地区：[1]北京航空航天大学机器人研究所,北京100083 [2]北京市创伤骨科研究所,北京100035

出　　处：《高技术通讯》2007年第9期919-923,共5页Chinese High Technology Letters

基　　金：国家杰出青年科学基金（60525314）、国家“十一五”支撑计划（2006BAI03A16）和北京市科技计划（H060720050230）资助项目.

摘　　要：针对C臂影像增强器采集的投影图像因失真变形而无法直接用于计算机辅助手术的问题,提出了一种基于多项式拟合的C臂投影全局校正法.该方法统一考虑C臂投影图像的3种类型的失真——针垫失真、S型扭曲、图像偏移,利用N阶多项式拟合图像的复合失真,然后利用最小二乘法求解最优化校正系数,从而具有所需标记点数量较少,校正区域连续,步骤简单,易于在线使用的优点.校正误差实验表明：该方法在3阶和4阶多项式的情况下所有标记点最大误差小于0.5像素,误差均方根小于0.26像素,并且对不同姿态的C臂投影图像的校正结果具有稳定性.该方法可用于计算机辅助手术导航、器械定位和术中X光三维锥束重建.In view of the fact that C-arm projection images acquired with the image intensifier are unsuitable for computer aided surgery because of image distortion, the paper proposed a polynomial fitting-based global correction method for C-ann projection. It considered three types of distortions, i.e., pincushion distortion, S distortion and image offset as a whole, and fitted the composite distortion of the images by the Nth order polynomial, and then obtained the optimum image correction coefficient by the least squares approach. With this method, fewer markers are needed and the correction regions are more continuous. The method is appropriate for online use in computer interventional procedures. Error experiments using projections of different typical poses were conducted. The results showed that the method＇ s maximal error of all markers was below 0.5 pixel and the root mean square error was below 0.26 pixel in the cases of the 3rd and 4th order polynomials. The global correction method can be used in the navigation, positioning and C-ann based cone-beam CT in computer assisted surgery.

关 键 词：图像失真 图像校正 误差分析 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]