基于B样条的医学图像弹性配准

为解决医学图像上的形变和偏移,本研究采用基于B样条自由形变的配准方法.建立一个由控制点组成的网格,并将其套用在待配准图像上,得到初始控制网格参数.以梯度下降法为优化算法,以SSD测度为相似性测度,不断修正该参数,参数的改变将引起控制点最近邻4×4控制网格内所有的像素点发生移动,移动后的坐标位置由B样条拟合得到.以双线性插值为插值算法,计算出该位置的灰度值,该位置相对于配准后图像,等同于像素点移动前的位置映射到参考图像相对应的位置,当配准后图像与参考图像对应位置的灰度差异度达到最小时,就达到图像的最佳配准.实验结果表明,该算法配准结果满意,从而验证了本研究算法的有效性.

Elastic Registration of Medical Image Using B-Spline

In order to solve deformation or deviation on medical image, Free-form Deformation (FFD) based on the B-Spline was used as the registration algorithm, which established a grid consisting of control points that is applied to the moving image to obtain the initial control grid parameters.Gradient descent measure was used as the optimize algorithm and sum of squared differences measure was used as the similarity measure to revise the parameters constantly.The change of parameters would cause pixel points in the nearest 4×4 unit control grids to move and the coordinate position after moving were gotten by B-Spline interpolation.Bilinear interpolation was used as the interpolation algorithm to calculate the grayscale value in the position.The position in the moved image relatively was equivalent to that in the fixed image mapped by the pixel point before moving.It would get best image registration result when the gray difference degree reached the minimum between the two corresponding positions respectively in the moved image and fixed image.Experiment results shows that it can get satisfactory registration result, thus verify the algorithm in this paper is effective.