在图像处理中常需要找出一幅图像与其对应的样板图像间的位移（ 变形） 量．本文比较了几种常用的图像匹配方法，并对一种基于梯度算子的弹性匹配方法进行了研究．这种方法首先构造一个二次误差指标函数．此指标函数由两项组成：匹配误差项和平滑约束项．此指标函数不同于Ｈｏｒｎ 和Ｓｃｈｕｎｃｋ 的光流法之处在于：其中的匹配误差项和平滑约束项直接定义于位移矢量场而不是光流场．利用变分法可以从这个二次误差指标函数导出一组椭圆形偏微分方程泊松方程（Ｐｏｉｓｓｏｎ ｅｑｕａｔｉｏｎ），这类方程有相当成熟的数值解法．采用有限差分法，对多组图像进行计算，证明此算法是有效的，并具有收敛速度快，鲁棒性好的特点 In image processing often need to find an image and its corresponding template image displacement (deformation) amount. This paper compares several commonly used image matching methods and studies a method of elastic matching based on gradient operator. This method first constructs a quadratic error index function. This indicator function consists of two items: matching error term and smoothing constraint term. The difference between the index function and Horn and Schunck’s optical flow method lies in that the matching error term and the smoothing constraint term are directly defined in the displacement vector field rather than the optical flow field. Using the variational method, a set of elliptic partial differential equations - Poisson equation can be derived from this quadratic error index function. These equations have well-established numerical solutions. The finite difference method is used to calculate multiple images, which proves that the algorithm is effective and has the characteristics of fast convergence and good robustness