大形变微分同胚图像配准快速算法

本文提出一种研究大形变图像配准算法. 大形变使得图像信息和拓扑结构有较大的改变, 目前该方面的研究仍然是一个难点. 基于严密数学理论的微分同胚Demons算法是图像配准的著名算法, 为解决大形变配准问题提供了重要基础. 基于对微分同胚Demons算法的研究结合流形学习的思想提出一种大形变图像配准的新算法(MRL算法). 新算法通过挖掘图像的局部和全局流形信息改进微分同胚Demons 速度场的更新, 更好地保持图像的拓扑结构. 对比实验结果表明, 本文所提出的算法能够快速高精度地实现大形变图像的配准.

A Fast Image Registration Algorithm for Diffeomorphic Image with Large Deformation

A registration algorithm for large deformation images is porposed. Since image information and topological structure undergo great changes with large deformation, image registration for large deformation images is a challenging work. The diffeomorphic demons algorithm, based on strict mathematical theory, is a famous image registration algorithm, which provides an important basis to solve the problem of large deformation image registration. Based on the study of the diffeomorphic demons algorithm, by combining the ideas of manifold learning, this paper presents a new algorithm for large deformation image registration (called MRL). The new proposed algorithm improves the diffeomorphic demons velocity field up by capturing both local and global manifold information of the image, and better maintains the topology of the image. Comparative experiment results show that the algorithm can quickly realize large deformation registration with a higher precision.