摘要
由于目前的视频图像拼接方法的计算较为复杂,导致特征点检测耗时较长,影响了视频图像拼接的效率,因此本文基于尺度不变特征变换(Scale Invariant Feature Transform,SIFT)设计了广电网络视频图像自动拼接方法。首先,通过广电网络视频图像采集与预处理,生成多尺度空间,重构图像信息;其次,限定图像重叠区域,归一化处理频谱,获取图像平移参数;再次,基于SIFT的特征提取与匹配,定位关键点,度量不同图像特征点的相似度;最后,视频图像自动拼接与融合,精炼变化矩阵,处理图像拼接缝合线。将本文设计的方法与两种传统方法进行对比实验。实验结果表明:本文方法的标定平均误差距离最小,具有更高的畸变校正精度,特征点提取的耗时最短,特征点实时检测效果更佳,拼接融合后的视频帧图像无拼接痕迹,未额外产生任何拼接误差,证明本文方法具有可行性。
Since the calculation of current video image stitching methods is complicated,which leads to long time consuming feature point detection and affects the efficiency of video image stitching,this paper designs an automatic video image stitching method for broadcasting networks based on Scale Invariant Feature Transform(SIFT).Firstly,through the video image acquisition and pre-processing of the broadcasting network,a multi-scale space is generated to reconstruct the image information;secondly,the image overlap region is limited,the normalized processing spectrum is normalized,and the image translation parameters are obtained;again,the feature extraction and matching based on SIFT locates the key points and measures the similarity of different image feature points;finally,the video image is automatically stitched and fused,the change matrix is refined,and the image stitching is processed stitching line.The method designed in this paper is compared with two traditional methods for experiments.The experimental results show that the method in this paper has the smallest calibration mean error distance,higher distortion correction accuracy,the shortest time consumption for feature point extraction,better feature point detection in real time,no stitching traces in the stitched and fused video frame images,and no additional stitching errors,which proves the feasibility of the method in this paper.
作者
常鑫
CHANG Xin(Tianjinshi Boyagshitong Technology Co.,Ltd.,Tianjin 300400,China)
出处
《信息与电脑》
2022年第11期173-175,共3页
Information & Computer
关键词
图像自动拼接
重叠区域
图像采集
广电网络视频图像
图像融合
image automatic mosaics
overlapping regions
image collection
video image of broadcast network
image fusion
