CAJ | 학술논문

提出了一种基于视频序列拼接的新方法。首先，利用KLT算法对视频序列中特征点进行提取和跟踪，实现关键帧粗略选取；其次，在选取的关键帧中利用SURF算法进行特征提取，利用最近邻距离比进行特征点匹配，通过RANSAC估计算法求精单映矩阵，并结合关键帧选取判定寻找最佳关键帧；最后，利用级联单映矩阵和加权融合算法实现视频序列拼接。实验验证了该方法的有效性。
제출료일충기우시빈서렬병접적신방법。수선，이용KLT산법대시빈서렬중특정점진행제취화근종，실현관건정조략선취；기차，재선취적관건정중이용SURF산법진행특정제취，이용최근린거리비진행특정점필배，통과RANSAC고계산법구정단영구진，병결합관건정선취판정심조최가관건정；최후，이용급련단영구진화가권융합산법실현시빈서렬병접。실험험증료해방법적유효성。
This paper proposes a new mosaic algorithm of video sequence. Firstly, the rough key frames are extracted from the video sequence using KLT feature extraction and tracking. Then, the features of the rough key frames extracted by SURF algorithm are matched by the nearest neighbor. With the RANSAC robust estimation for homography matrix and key frames selection judgment, the final key frames are extracted. Finally, with the cascade of homography matrix, the seamless mosaic is rendered by the image blending. The experiment demonstrates this algorithm＇s efficient.