It is time consuming for surface registration of medical volumetric images in high resolution.Fully automatic methods cannot handle complex situations well,for example,massive non-continuous holes in children's skull.To address this problem,we propose an interactive registration technique based on an efficient deformation algorithm.We first sample the volumetric data;combine local as-rigid-as-possible energy with non-rigid iterative closest point as an automatic registration framework and integrate real-time interactive refinement in local regions.The experiments demonstrate our technique is able to complete the registration in 13 seconds for an average 1.2 million sample points with the same accuracy by marching cubes.%对高分辨率体数据构成的医学图像进行隐式曲面配准是一件耗时的工作,对于发育未完全的儿童头骨中包含的大量不连续空洞这样的复杂情况,全自动算法一般难以处理,为此提出一种交互式的快速配准方法.首先对一定范围内体数据进行采样得到目标点集;而后将定义在模板网格上的局部最刚性变换能量引入非刚性最近点迭代配准中作为自动配准框架;在此基础上,加入用户实时交互对局部区域结果进行调整与优化.实验结果表明,对于平均120万体素采样点,该方法能够在13s内完成配准过程,并且与marching cubes结果具有相似准确度.
展开▼
