基于改进快速鲁棒特征的实时视频拼接研究  

Real-time Video Stitching Based on Improved Speeded Up Robust Feature Algorithm

作　　者：闫岩[1] 李杨[1] 

机构地区：[1]91245部队,葫芦岛125000

出　　处：《计算机与数字工程》2017年第4期727-730,750,共5页Computer & Digital Engineering

基　　金：国家自然科学基金:广义测不准原理及其应用研究(编号:61002052);基于调制解调的图像多分辨率分解方法(编号:61273262)资助

摘　　要：针对视频拼接对实时性要求较高的实际,提出了基于SURF算法的改进图像配准算法。首先匹配视频首帧图像并限定特征点的提取区域,其次采用最近邻和次近邻算法完成图像特征的粗匹配,最后通过最外侧匹配点检测图像重叠区域,并在该区域内完成后续视频帧的计算。仿真结果表明,改进的SURF算法显著降低了运算量、提高了匹配速度,可以完成实时视频的生成。This article presents an improved image registration algorithm that is based on SURF to solve the real-time problem in video stitching area.First,it registers the first frame of video image and restricts the extraction area of feature points.Then it matches the feature points based on the nearest neighbor and second neighbor algorithm.At last,it finishes the calculations about the subsequent frames of the image overlapping area which is detected by the lateral matching points.

关 键 词：视频拼接 SURF 图像配准 图像融合 

分 类 号：TP301[自动化与计算机技术—计算机系统结构] TP751[自动化与计算机技术—计算机科学与技术]