经典图像快速匹配算法比较分析研究  

Comparative Analysis of Classical Image Fast Matching Algorithms

作　　者：王坚[1] WANG Jian(School of International Business and Economics,Liaoning University,Dalian 116000,Liaoning)

机构地区：[1]辽宁对外经贸学院,辽宁大连116000

出　　处：《攀枝花学院学报》2022年第5期79-85,共7页Journal of Panzhihua University

基　　金：2021年度辽宁省普通高等教育本科教改研究项目“应用型高校学生数字化素养提升研究与实践”(2021SJJGYB13)。

摘　　要：为满足图像匹配效果愈来愈高的要求,剖析不同匹配方式的图像处理技术,筛选出图像快速匹配算法。分析比较基于误差平方和(SSD)、改进基于绝对误差平方和的序贯相似性算法(SSDA)和基于灰度编码(PFC)的算法的基本原理与匹配方式。在MATLAB平台上,以相同灰度图像模板为研究对象,对三种经典算法和基于学习的不变特征变换算法进行图像匹配运算时长比较。结果表明,按匹配时长由大到小排序分别为:SSD>IFT>SSDA>PFC。在经典算法中,局部灰度编码算法实现了快速匹配目标。相较于经典算法,LIFT算法引入SIFT特征向量,提高经典算法的匹配性能,但运行时间延长。In order to meet the increasingly higher requirements of image matching effect,the image processing technology of different matching methods is analyzed,and the fast image matching algorithm is selected in order to analyze and compare the basic principles and matching methods of the algorithm based on the sum of squares of errors(SSD),the improved sequential similarity algorithm based on the sum of squares of absolute errors(SSDA)and the algorithm based on grayscale coding(PFC).On the MATLAB platform,the image template of the same gray scale is taken as the research object,and the image matching operation time is compared among the three classical algorithms and the learning-based invariant feature transformation algorithm.The results show that the order of matching duration is SSD>LIFT>SSDA>PFC.In the classical algorithm,the local gray coding algorithm achieves fast target matching.Compared with the classical algorithm,LIFT algorithm introduces SIFT feature vector and improves the matching performance of the classical algorithm,but results in longer elapsed time.

关 键 词：图像匹配 绝对误差和 灰度编码 MATLAB 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]