关键词: 图像配准, 链码, 不变矩, 仿射变换

Abstract: In the traditional image registration method which is based on the feature of chain code，there are many drawbacks like high computational complexity，hard to extract boundary，not ideal accuracy and so on.This paper presents a new algorithm for image registration which is based on the features of patches and chain code.First，with patch detection，the closed boundary regions in both images are extracted and the computational complexity is reduced.Next，the correspondence of closed boundary regions is developed by chain code correlation and invariant moments.Finally，the center of gravity in the correspondent regions is used as matching points and a consistency check is applied.The affine transformation parameter is estimated by matching points.Experimental results show that the method is fast and robust，and the registration accuracy is better.

Key words: image registration, chain code, moment invariants, affine transformation