一种改进的宽基线图像SURF算法  

Improved Wide Baseline Image SURF Algorithm

作　　者：于海燕[1] 姚青华[1] 

机构地区：[1]郑州科技学院信息工程学院,河南郑州450064

出　　处：《电视技术》2015年第21期28-29,39,共3页Video Engineering

基　　金：郑州市科技局自然科学项目(20130744)

摘　　要：通过对SURF特征点匹配过程的研究,提出一种改进的宽基线图像SURF算法。提取SURF特征点,在特征点匹配过程中采用改进方法计算图像间的尺寸、旋转和平移参数,减少匹配点搜索范围,运用RANSAC算法剔除误匹配点对。实验表明该算法在图像间有较大差别时降低了误匹配率,具有良好的实时性,达到了较好的应用效果。An improved wide baseline image SURF algorithm is proposed based on the research of SURF feature point matching process.Extracts SURF feature points, then, rotation ,scale and shift parameter of images are estimated by Fourier-Mellin transform in order to reduce search scope in the process of the feature points matching, more precise matching points could be obtained through the RANSAC algorithm. Experiments prove this algorithm has lower false match rate and good real-time ability is obtained by this algorition when larger difference between images, a better application effect has achieved.

关 键 词：SURF算法 Fourier—Mellin变换 宽基线图像 点匹配算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]