一种改进的SIFT特征点匹配算法  

Improved SIFT Algorithm for Feature Point Matching

作　　者：郭振成[1] 梁凤梅[1] 

机构地区：[1]太原理工大学信息工程学院,山西太原030024

出　　处：《电视技术》2014年第1期22-25,119,共5页Video Engineering

摘　　要：提出了一种改进SIFT特征点匹配算法,旨在提高图像的特征点匹配算法效率。SIFT特征点检测算法检测到的特征点数量较大,且每个特征点描述子都是128维的向量,算法匹配时间较长,实时性较差,改进的算法将特征点描述子的维数从128维降低到48维,而描述子像素范围从16×16提高到24×24。实验表明,改进的算法不仅降低了匹配时间,而且提高了匹配精度。In order to improve the efficiency of feature point matching algorithm,an improved SIFF feature point matching algorithm of the image is pro- posed. For the amount of feature point detected by SIFT feature point detection algorithm is large and each feature point descriptor is 128 dimensional vector, matching time of the algorithm is long. In this paper,dimension of the feature, descriptor described from 128d down to 48d,and the range from 16 × 16 up to 24 x24,experimental results show that the improved algorithm can not only improve the matching speed but also improve the matching accuracy.

关 键 词：SIFT算法 特征描述子 特征点检测 匹配 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]