基于边缘几何特征的图像精确匹配方法

提出一组快速高精度计算切线斜率的五点公式，用以估计图像边缘曲线的角度特征，并利用角度直方图估计图像几何变换的旋转参数，实现具有大旋转差异图像间的粗匹配．在进行角度补偿后，利用灰度互相关判据搜索匹配点对，计算出几何变换参数，实现较高精度的旋转和平移校正，最后用松弛迭代法完成图像的精确匹配．与基于小波方向角特征的匹配方法相比，文中方法利用图像中主要的边缘信息实施匹配，具有较好的鲁捧性，可成功实现对各类具有较大相关程度图像间的精确匹配，对图形匹配也具有重要意义．

A Precise Image Registration Method by Utilizing the Geometric Feature of Edges

In this paper, a new image registration algorithm is proposed to perfectly address the problem caused by rotation between images Our algorithm is based on the observation that the geometric features at the edge of an image can provide important information for the image registration The main novelty of the algorithm is in proposal of a fast high precision formula being applied to calculate the tangent angles of any point at the edges, and these angles are used as robust features to estimate the approximate rotation angle between two images By the approach, we first detect the edge of the original images for a set of feature points, calculate their tangent angle with our new formula and then use the statistical differences of the tangent angles of all feature points within two images to estimate the rotation parameters of the geometric transform Finally, a step of coarse to fine image registration is performed based on the correlation criterion Experimental results in various types of images have been provided to verify the accuracy of our automatic registration algorithm