摘要
针对基于加速鲁棒特征(SURF)的图像配准算法的配准精度易受重复特征干扰的影响这一问题,提出一种基于SURF的抗重复特征干扰的图像配准方法。使用SURF算法提取图像特征点;针对重复特征干扰,提出一种特征点分类匹配方法以取代传统的全局匹配,在不显著增加计算量的情况下有效的降低误配率;使用随机抽样一致(RANSAC)算法进一步筛除误配,并计算出图像转化矩阵以完成配准。实验结果表明,该方法能有效抑制实验图像中严重的重复特征干扰,并获得较高的配准精度。
Aiming at the problem of the registration accuracy in SURFbased image registration algorithm is easy to he affected by the interference of repeated features, a SURFbased image registration method with antiinterference of repeated features is proposed. Firstly, feature points of image are extracted with SURF algorithm. Secondly, aiming at the repetitive feature inter ference, a feature point classified matching method is put forward to replace the traditional overall matching, which would effec tively lower the mismatching rate without increasing the calculated amount. Finally, RANSAC algorithm is adopted for further elimination of mismatch and calculation of the image transition matrix to finish registration. The experiment results shows that severe repeated feature interference in the experimental image can he controlled effectively with relatively high registration preci sion by employing this method.
出处
《计算机工程与设计》
CSCD
北大核心
2013年第3期978-982,共5页
Computer Engineering and Design
基金
江苏省产学研前瞻性联合研究基金项目(BY2012027)
江苏省六大人才高峰高层次人才基金项目(JXQC-139-132)
