基于边缘的图象配准改进算法

现有的图象配准方法对旋转处理比较困难，且处理效率不高。本文提出一种基于边缘的图象配准改进算法，该算法首先用小波的快速方法提取边缘特征点，通过线拟合模型消除噪声；然后得出特征点的边缘方向，定义“角度直方图”来估计两幅待配准图象间的旋转角，应用线性加权方法消除奇异点，最后得到图象间的精确变换关系。

Image Registration Using Edge-Based Methods

In this paper,an improved edge based method for image registration is proposed by using the accelerated wavelet transform to extract feature points as the basis for registration.All the edge directions of the feature points are estimated based on a line fitting model.The concept of angle histogram is defined to decide the rotation angle.During real matching,we check each candidate pair in advance to see if it can possibly become a real matching pair,then we use linear weights to extract correct matching pairs.