CAJ | 학술논문

针对经典弹性配准方法在医学图像应用上计算复杂、方向信息不足的问题进行了研究，提出一种基于非下采样剪切波变换（NSSTR）的医学图像配准方法。利用剪切波能够对各个尺度、方向和位置实现较好定位的优势，先用该算法对医学图像作非下采样剪切波变换，得到各尺度、方向子带的剪切系数，然后对高频、低频的变换图像分别使用不同的配准方法，最后统一到变化网格上得到最终配准图像。实验结果表明，本方法与传统方法相比，不仅具有配准精度高、鲁棒性好的特点，而且计算效率更高。
침대경전탄성배준방법재의학도상응용상계산복잡、방향신식불족적문제진행료연구，제출일충기우비하채양전절파변환（NSSTR）적의학도상배준방법。이용전절파능구대각개척도、방향화위치실현교호정위적우세，선용해산법대의학도상작비하채양전절파변환，득도각척도、방향자대적전절계수，연후대고빈、저빈적변환도상분별사용불동적배준방법，최후통일도변화망격상득도최종배준도상。실험결과표명，본방법여전통방법상비，불부구유배준정도고、로봉성호적특점，이차계산효솔경고。
The computational complexity of medical imaging and insufficient direction information are the drawbacks of classi-cal elastic registration methods.Based on the study for the problem,this paper proposed a new medical image registration algo-rithm based on non-subsampled shearlet transform(NSSTR).In the method,it used the privilege that shearlet could locate each scale,direction and position.Firstly,the algorithm used non-subsampled shearlet transform on medical image to obtain scale,di-rection of subband shear coefficient,then adopted different registration methods for high frequency and low frequency images respectively.Finally,the algorithm got changing grid and registration image.Experiments show that the new proposed method outperforms traditional methods in the aspect of registration precision and robustness.Especially the algorithm is with higher computational efficiency.