引入边缘信息度量的自由形变图像配准  

Free-form deformation image registration using edge information measure

作　　者：于慧敏[1] 廖秀秀[1] 

机构地区：[1]浙江大学信息与电子工程学系,浙江杭州310027

出　　处：《浙江大学学报（工学版）》2008年第10期1700-1705,共6页Journal of Zhejiang University：Engineering Science

基　　金：浙江省科技计划资助项目(2006C21035)

摘　　要：针对形变图像配准问题,提出一种新的图像配准算法.配准算法采用基于B样条的多分辨率网格自由形变(FFD)的图像形变建模,用于解决局部形变的图像配准问题.为了改善局部形变图像的配准精度,在MSD度量标准中引入边缘配准测度,提出了一种改进的图像相似性度量方法,强调了边缘信息对配准的贡献.为了解决形变图像的局部配准问题,设计了一种可选择的交互式配准方法,用于图像配准后,进一步减少局部区域的配准误差.实验结果验证了本文方法有很好的配准精度,得到平滑、连续且一对一映射的变换域.A novel algorithm was presented aiming at the problem of deformation image registration. A Bspline based free-form deformation （FFD） model with multi-resolution grid was established to resolve the problem of local deformation image registration. An improved method for image similarity measure was proposed to improve the accuracy of local deformation image registration. The method introduced edge registration measure into the mean of squared differences （MSD） criterion and emphasized the contribution of edge information for registration. An optional interactive registration was suggested to further decrease the error for local regions with serious distortion after image registration to resolve the problem of local registration. Experimental results show that the algorithm has a satisfactory registration precision and the recovered transformation field is smooth, continuous and one-to-one mapping.

关 键 词：形变图像配准 自由形变 边缘信息 交互式配准 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]