图像配准的复制动力学方法

Replicator equation based image registration approach

为了实现高精度图像配准,把单个像素看做参与博弈的局中人,将图像配准视为图像中各个像素与其周边像素的演化博弈过程。为了寻求博弈的均衡点,根据配准问题的特点,对复制方程进行修改,提出基于修改后的复制方程的图像配准方法。新方法是完全基于问题域建立起来的,能实现对任意形变的高精度配准,并且配准精度与分辨率无关。针对刚性形变的遥感图像和非刚性形变的医学图像进行了两组实验,结果表明了新算法的有效性。

If model a pixel as a player,then image registration can be viewed as an evolutionary game,in which each pixel play with its neighbors.To find an equilibrium point of the evolutionary game,this paper adapted the well known replicator equation according to characteristics of image registration,and proposed a registration algorithm based on the adapted replicator equation.It developed the novel algorithm on problem domain totally,it could register any complex deformation with high accuracy,and furthermore the accuracy was robust to image resolution.Experimentations on rigid-deformed remote sensing images and nonrigid-deformed medical images are designed,the results show that the proposed approach is effective.