为消除视频序列之间的冗余信息，以简单的摘要形式表达视频的主要内容，提出了一种基于视频序列的图像拼接方法。首先，采用改进的帧间聚类算法提取视频的关键帧；其次，利用SIFT算法提取关键帧的特征点，采用最近邻算法进行特征点匹配，通过引导互匹配法和投票过滤法提高匹配精度；再次通过RANSAC鲁棒估计算法得到所选帧间的单映矩阵，并使用LM非线性迭代算法对单映矩阵进行精炼。最后，利用级联单映矩阵结合加权融合算法实现了视频序列的无缝拼接，实验效果较为理想。%In order to eliminate the redundant information between video sequences, express the main content of the video in a simple abstract form, an image mosaic method based on video sequence was proposed. First of all, the key frames were extracted using the improved inter frame clustering algorithm. Secondly, the feature of key frames were extracted by using the SIFT algorithm, the nearest neighbor algorithm is used to match the feature points, guided complementary matching and voting filter is used to improve the matching accuracy. Then with the RANSAC robust estimation algorithm to get the selected frames' homography matrix. The LM nonlinear iteration algorithm was used to refine the homography matrix. Finally, cascade homography matrix and weighted fusion algorithm is used to realize the seamless splicing of video sequences. The experiment demonstrates this method is efficient.
展开▼
