一种改进的快速图像拼接方法  

AN IMPROVED FAST IMAGE MOSAIC METHOD

作　　者：曲天伟[1] 陈晓丹[2] 曹雪栋[1] 陈桂兰[2] 

机构地区：[1]黑龙江工程学院计算机科学与技术系,黑龙江哈尔滨150050 [2]黑龙江科技学院计算机与信息工程学院,黑龙江哈尔滨150027

出　　处：《计算机应用与软件》2011年第7期136-140,共5页Computer Applications and Software

基　　金：黑龙江省教育厅科学技术研究项目(12511447)

摘　　要：为了提高图像拼接的速度,提出了一种快速的图像拼接方法。首先在SUSAN角点检测算法检测出图像角点的基础上,采用图像分块和邻近角点剔除的方法来保证图像角点分布均匀并且避免出现角点聚簇现象,利于提高拼接的精度。其次,利用灰度相关性进行特征角点的匹配并消除伪匹配。然后采用改进的RANSAC算法快速地估计变换矩阵,该算法中采用预检测的方法快速抛弃那些不是候选模型的临时模型,加快了算法的速度。最后进行颜色融合,生成无缝拼接图像。实验结果表明,该方法在得到较高精度的情况下,大大减少了运算量,提高了图像拼接的速度。To improve image mosaic speed,a fast image mosaic method is proposed.Firstly on the basis of detecting image corners by SUSAN corner detection algorithm,the method uses image breaking and neighboring corner elimination method to assure the image corners are evenly distributed while avoiding corner gatherings,which helps improving mosaic accuracy.Secondly with gray relativity the characteristics corners are matched to eliminate forgery matches.Then the improved RANSAC algorithm is used for fast estimation of transform matrix.In this algorithm the pre-detection method is used to quickly discard temporary models which are not candidate models so that the speed of the algorithm is upgraded.Finally color fusion is executed to generate seamless mosaic images.Experiment results show the method acquires higher precision,carries out less calculation,and consumes less time on image mosaic.

关 键 词：角点检测 角点匹配 随机抽样一致性算法 图像拼接 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]