基于仿射几何不变量的鲁棒图像配准算法

Robust Image Registration Method Based on Affine Geometric Invariant

提出了一种新的基于特征的鲁棒图像配准算法.该算法通过构造特征区域将仿射几何不变量应用到特征描述过程中,使其能够处理待配准图像之间满足仿射几何畸变的情况.由于在特征描述过程中,特征矢量计算量小且仅与控制点的坐标信息有关而独立于图像的灰度信息,故该算法不仅具有很好的时效性,而且对图像之间灰度值的偏差具有很好的鲁棒性.

A new robust feature-based image registration algorithm was presented.This algorithm employs the affine geometric invariant in the design of feature descriptor by constructing feature areas,which enables it to deal with affine geometric deformation.Since the feature descriptor,which has a small amount of calculation,is only determined by coordinate information and independent of the gray scale information,this method has good timeliness and is robust to the gray scale discrepancy between the images.