摘要
针对SIFT算法在进行图像配准时存在提取特征点数目大、无法精确控制、运算速度慢、配准点精度不高的问题,提出一种基于局部显著特征的快速图像配准方法。该方法首先对原始图像和待配准图像进行降采样,对降采样图像分别提取SIFT特征点,并对特征点运用改进的K-means聚类算法进行聚类;然后利用聚类结果筛选聚类区域,在各聚类区域提取显著特征点进行粗匹配;最后利用显著特征点在原始图像中定位显著区域,对所得显著区域进行精配准。实验结果表明,该方法减少了图像匹配时间,控制了特征点数量,在保证匹配准确度的同时,有效地提高了特征匹配的效率。
Aiming at the problems such as large number of feature points,difficulty in accurate control,low computation speed and registration accuracy in SIFT algorithm,this paper presented a fast image registration method based on local salient feature.In the algorithm,it firstly down sampled the original input and reference images to reduce spatial resolution,then extracted the SIFT feature points from the down sampled images and clustered by the improved K-means algorithm.Subsequently,it obtained the clustered regions by filtering the above clustered results,extracted and matched the salient feature points coarsely in the clustered regions.Finally,it located the salient regions according to the matched salient point pairs and matched to achieve accurate registration.Experiment results show that the method can reduce the matching time,control the number of the feature points,and also improve the efficiency significantly when keeping a higher matching precision.
出处
《计算机应用研究》
CSCD
北大核心
2012年第11期4370-4374,共5页
Application Research of Computers
基金
陕西省自然科学基金资助项目(2010JM8014)
中国博士后科学基金资助项目(20100471838)
中国博士后特别基金资助项目(201104787)
