基于SIFT特征和角度相对距离的图像配准算法  

Relative Angle Distance for Image Registration Based on SIFT Feature

作　　者：李晖晖[1] 郑平[1] 杨宁[1] 胡秀华[1] 

机构地区：[1]西北工业大学自动化学院,陕西西安710129

出　　处：《西北工业大学学报》2017年第2期280-285,共6页Journal of Northwestern Polytechnical University

基　　金：航空科学基金(20131953022)资助

摘　　要：对2幅不同角度、不同光照条件或不同相机采集到的图像进行配准,是一项十分具有挑战性的研究。针对参考图像和待配准图像对之间存在的仿射变换问题,提出了一种灵活通用的、基于SIFT特征和角度相对距离的图像配准算法。算法充分利用了图像正确匹配特征点对之间存在的角度关系,实现了特征点之间的精确匹配。将所提算法同LLT(locally linear transforming)算法及RANSAC算法进行了对比实验,结果表明,新算法有较高的有效性和鲁棒性。而且新算法不仅适用于普通图像,在近红外与可见光图像以及遥感图像中均充分体现了良好的鲁棒性和适用性,在匹配特征点数目较少时,也具有良好的鲁棒性。Registration between images taken from different cameras, from different angles or under different lighting conditions is a challenging research. To solve the affine transformation existed between the reference image and the image to be registered, this paper presents a flexible and robust image registration algorithm based on SIFI＂ and the relative angle distance. This algorithm takes full advantage of the relationship of angle between the correct matching feature points and effectively screens out the corrected matching points by the relative angle distance. Mo- reover, this algorithm can get the corrected matching points only use the SIFT features simply without any other in- formations. Compared with the algorithm such as LLE（ Locally Linear Transformation） and RANSAC, this algorithm can get more corrected matching points in most cases. This algorithm can apply not only to ordinary image, but also to infrared and visible images, remote sensing images. This algorithm has good robustness even if small number of matching features existed.

关 键 词：图像配准 SIFT特征 角度相对距离 精确匹配 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]