首页> 外文会议>International Conference on Smart City and Systems Engineering >A Novel Image Registration Algorithm Using SIFT Feature Descriptors
【24h】

A Novel Image Registration Algorithm Using SIFT Feature Descriptors

机译:一种使用SIFT特征描述符的新型图像配准算法

获取原文

摘要

This paper focuses on the problem of image registration, which is a fundamental and important problem in computer vision. As local features are effective to tackle the task of image registration, hence, we exploit SIFT feature descriptors to describe the interesting key points in images. The reasons is that SIFT can extract features which are invariant to scaling, orientation, affine transforms and illumination changes. Afterwards, to lower the effects of outliers, the nearest distance ratio method is used to get initial matched feature descriptor pairs, and then scale-orientation joint restriction is used to detect false matching SIFT descriptor pairs. After using Random Sample Consensus to prune outliers, image registration results can be obtained. In the end, we develop an experiment based on QuickBird and IKONOS datasets, and very positive results are achieved.
机译:本文重点介绍了图像登记的问题,这是计算机视觉中的基本和重要问题。随着本地特征对于解决图像登记的任务而有效,因此,我们利用SIFT特征描述符来描述图像中的有趣关键点。原因是SIFT可以提取不变的特征,以缩放,方向,仿射变换和照明变化。之后,为了降低异常值的影响,最接近的距离比方法用于获得初始匹配的特征描述符对,然后使用尺度方向接合限制来检测假匹配的SIFT描述符对。在使用随机样本的共识后,可以获得图像登记结果。最后,我们开发了基于Quickbird和Ikonos数据集的实验,实现了非常积极的结果。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号