基于非刚体配准的体数据插值重建方法

Nonrigid Registration Based Interpolation Approach

三维图像的处理和操作需要将一般的断层序列插值成为具有各坐标轴一致的分辨率的体数据,而目前最常用的线性插值方法在层间距较大时会导致图像边缘模糊和出现伪影。Penney根据现有的非刚体匹配方法,提出了利用图像形变场数据的插值算法,大大提高了层间插值的质量。本文对Penney提出的算法进行了两方面的改进,在配准过程中用简单的单射性约束取代了复杂的平滑性约束,用邻域平均算法替代Penney使用的最邻近直线插值方法,并将新算法的实验结果与原算法、线性插值进行了对比。新算法在保持高质量插值的前提下提高了计算速度。该算法可以应用于精度要求比较高的体数据插值重建过程。

For common medical slices（e.g.CT or MR）,the space between neighboring pixels in a slice is often smaller than the center-to-center slice separation, while in most 3D image processing, analysis and visualization requires voxels to be isotropic. Many interpolation techniques have been proposed for processing such data. The commonly used grey level interpolation will produce artifacts or blurred contours when two given slices shift considerably. Shape-based methods achieve more reasonable interpolation results, especially the nonrigid registration based methods. This paper takes modified Rueckert＇s registration method to produce deformed data fields, and uses simple neighbor averaging method to interpolate scattered data. Compares to Penney＇s method, this one produces almost the same result image with less computation time.