图像配准中B样条滤波和GPVE的融合插值算法  

Mixed interpolation algorithm integrating B-spline filter and generalized partial volume estimation in image registration

作　　者：胡顺波[1] 

机构地区：[1]临沂大学信息学院,山东临沂276005

出　　处：《计算机应用》2011年第8期2225-2228,共4页journal of Computer Applications

基　　金：山东省优秀中青年科学家科研奖励基金资助项目(BS2009DX028);山东省高校科研发展计划项目(J09LG82)

摘　　要：针对B样条GPVE插值法和B样条滤波法产生归一化互信息(NMI)测度伪极值点的原因,结合它们导致联合直方图聚散程度的互补效果,提出了各阶B样条滤波和GPVE的融合插值算法。通过图像之间的刚体配准实验,从测度曲线光滑性能和极值点数目方面,对比分析了一阶、三阶、五阶B样条滤波法,B样条GPVE插值法和新提出的融合插值算法。实验结果表明,提出的各阶B样条融合插值算法的配准性能都优于对应阶次B样条滤波法和B样条GPVE插值法。Many false maxima were generated by B-spline Generalized Partial Volume Estimation（GPVE） or B-spline filter,but they had opposite dispersion effects on joint histogram of two registering images.By integrating the same order B-spline filter and B-spline GPVE,the mixed interpolation algorithm was proposed.When the registered images were rigidly aligned,the smoothness and number of maxima of the normalized mutual information curves were compared by applying many interpolation algorithms,which included the order 1,3,and 5 B-spline filters,B-spline GPVE,and the new mixed interpolation algorithms.The experimental results show that the new mixed interpolation algorithms with order n B-spline outperform the corresponding B-spline filter and B-spline GPVE in image registration.

关 键 词：图像配准 归一化互信息 插值 B样条 

分 类 号：TP391.413[自动化与计算机技术—计算机应用技术]