基于双向异步投票策略医学图像特征点匹配  

Medical image feature point matching based on two-way asynchronous voting strategy

作　　者：王双玲 龙成章 刘辉 邓振生[2] 

机构地区：[1]河南省计量科学研究院,河南郑州450008 [2]中南大学信息物理工程学院,湖南长沙410083

出　　处：《中国医学物理学杂志》2016年第4期423-426,共4页Chinese Journal of Medical Physics

摘　　要：图像特征点匹配作为图像处理领域的一项基本技术,在临床医学图像处理中具有很重要的意义。本文根据医学图像的特点提出一种基于双向异步投票策略的特征点匹配方法。首先,利用改进的harris角点检测算子提取图像角点作为图像的特征点;其次,采用本文提出的基于双向异步的投票策略的角点匹配算法完成角点的粗匹配,使待拼接图像的特征点大致对应;最后利用RANSAC优化算法剔除误匹配,使特征点达到精确匹配。Image feature point matching, a basic image processing technology, is significantly important in the clinical image processing. According to the characteristics of medical images, a feature point matching method based on two- way asynchronous voting strategy was proposed. Image corners, feature points of the image, were firstly extracted by the improved harris corner detection; the proposed corner matching algorithm based on two-way asynchronous voting strategy was applied to complete the rough matching of corner points, making the feature points of the stitching image roughly corresponding; finally, the false matching was eliminated by RANSAC optimization algorithm to accurately match the feature points.

关 键 词：医学图像 特征点匹配 双向异步 投票策略 图像拼接 

分 类 号：R319[医药卫生—基础医学]