基于相位相关法与改进SURF算法的图像拼接方法  

Image Stitching Method Based on Phase Correlation and Improved SURF

作　　者：刘鹏飞 高如新[1] LIU Peng-fei;GAO Ru-xin(Institute of Electrical Engineering and Automation,Henan Politechnic University,Jiaozuo 454000,China)

机构地区：[1]河南理工大学电气工程与自动化学院

出　　处：《软件导刊》2019年第11期157-160,164,共5页Software Guide

基　　金：河南省自然科学基金项目(162300410126);河南省科技攻关项目(162102210228);河南省高校基本科研业务费专项资金资助项目(NSFRF1617)

摘　　要：为避免相邻图像非重叠区域特征点被检测和提取,加速图像配准,提出一种基于相位相关法与改进SURF算法的快速图像拼接方法。该方法采用相位相关算法确定待拼接图像的重叠区域,限定SURF特征点检测、提取范围,用改进的SURF算法进行特征点匹配;然后根据MSAC算法剔除误配后的特征点匹配对,求取单应性矩阵,实现图像之间的快速配准;最后采用多波段融合算法对配准后的图像进行处理以消除拼接缝。实验结果表明,与传统算法相比,该方法可提高特征点匹配正确率,加速图像配准过程,完成拼接图像间的平滑过渡。In order to avoid detection and extraction of feature points in non-overlapping regions of adjacent images and accelerate image registration,a fast image mosaic method based on phase correlation and SURF improvement was proposed.The innovation of this method lies in the use of phase correlation algorithm to determine the overlapping areas of images to be spliced,and limit the detection and extraction range of surf feature points.Surf algorithm was improved for feature point matching.Then the mismatched feature point pairs were eliminated by MSAC algorithm to obtain the homography matrix.Multi-band blending algorithm is used to process the registered image and eliminate the seam.Experimental results show that compared with the traditional algorithm,this method improves the accuracy of feature point matching,accelerates the image registration process,and completes the smooth transition between the splicing images.

关 键 词：图像配准 相位相关法 改进SURF算法 特征点匹配 图像融合 多波段融合 

分 类 号：TP317.4[自动化与计算机技术—计算机软件与理论]