基于块信息的异模图像配准算法

摘要

红外图像和可见光图像的配准是一类极为典型的异模配准.为了解决两者在配准速度和配准精度上的问题,在建立积极点、消极点、积极信息和消极信息4种概念的基础上,提出了一种具有高度鲁棒性的红外和可见光图像配准新方法.对参考图像和浮动图像进行块信息的提取,并对提取出的块信息进行重合度相似性检验,最终实现红外和可见光图像的高精度配准.实验证实本项研究成果具有较强的抗噪能力,配准精度可以达到亚像素级.%The registration of infrared images and visible images is very representative in different modality image registration.In order to improve the accuracy of the registration,four new concepts were proposed,including positive point,negative point,positive information and negative information.Depend on this four new concepts,a new algorithm of high robustness was developed.The algorithm uses region the growing method to extract the block information of images,and matches the block images based on the contact ratio to realize the high accurate registration.The experimental results demonstrate that the algorithm has good noise immunity performance,and can get sub-pixel registration precision.