CAJ | 학술논문

本文采用微分同胚变换预处理图像，得到初始化形变场，提高对形变图像的配准精度；采用Broyden族算法优化能量函数，自动确定迭代次数，提高优化效率；基于Demons算法思想引入图像梯度灰度场相似量构造能量函数，提高灰度信息少的图像配准精度。实验证明，本文算法配准精度优于改进的Demons算法，尤其在配准大形变图像时，本文算法配准精度高的优势更加明显。
본문채용미분동배변환예처리도상，득도초시화형변장，제고대형변도상적배준정도；채용Broyden족산법우화능량함수，자동학정질대차수，제고우화효솔；기우Demons산법사상인입도상제도회도장상사량구조능량함수，제고회도신식소적도상배준정도。실험증명，본문산법배준정도우우개진적Demons산법，우기재배준대형변도상시，본문산법배준정도고적우세경가명현。
To improve the accuracy of the deformable image registration,we utilize diffeomorphism transforma-tion to pretreat the image and get the initialized deformation field.to determine the number of iterations automatically and improve the efficiency of optimization,we utilize broyden family algorithm to optimize the energy function.to in-crease the accuracy of the little gray information image registration,we introduce the Image gradient gray field to construct the new energy function.Our experiments show that the accuracy of this paper algorithm is better than the improved Demons algorithm.