CAJ | 학술논문

在视频拼接过程中，需要进行大量的图像处理计算，高复杂度的拼接算法很难满足实时性要求。因此，为了实现多视频源快速拼接，不仅要求拼接算法具有较强的鲁棒性，同时还需要具有较低的复杂度。文中提出了一种新的快速宽视频拼接方法，此方法首先利用SURF算子提取图像特征点并进行匹配，接着使用多分辨率融合算法进行全景图融合，然后利用基于路径的方法结合光流实现低复杂度的视频插帧，同时整个宽视频拼接过程使用CUDA进行并行加速，最后用一个双目拼接系统对文中提出的方法进行可行性验证。
재시빈병접과정중，수요진행대량적도상처리계산，고복잡도적병접산법흔난만족실시성요구。인차，위료실현다시빈원쾌속병접，불부요구병접산법구유교강적로봉성，동시환수요구유교저적복잡도。문중제출료일충신적쾌속관시빈병접방법，차방법수선이용SURF산자제취도상특정점병진행필배，접착사용다분변솔융합산법진행전경도융합，연후이용기우로경적방법결합광류실현저복잡도적시빈삽정，동시정개관시빈병접과정사용CUDA진행병행가속，최후용일개쌍목병접계통대문중제출적방법진행가행성험증。
In the process of video stitching,high complexity algorithm can't use to real-time video stitching since most image process contains mass of computing. In order to achieve real-time stitching,the stitching algorithm must have strong robustness and lower com-plexity. In this paper,propose a new method about fasting-video stitching. First,it extracts and matches the image feature points by using SURF,and blends the panorama image by using multi-resolution fusion algorithm. Then achieve a low complex video interpolation by combining the base-path with optical flow. Meanwhile it accelerates the whole process by using CUDA. At last,a binocular stitching sys-tem is introduced to verify the whole process's feasibility.