一种基于压缩域的视频拼接算法

Video mosaic algorithm for compressed video

提出了一种基于压缩域的双摄像机视频拼接算法。首先利用相位相关法估算输入视频的对应第一帧重叠区域,并在重叠区域进行SIFT角点检测和匹配,加快角点匹配速度,提高匹配稳健性,使用RANSAC算法去除外点,采用奇异值分解法配合LM非线性优化方法求解变换参数,得到首帧的对应投影矩阵;对于非首帧的配准,利用压缩视频中的当前帧与前帧的运动矢量,获得全局运动矢量,然后结合对应前帧的投影矩阵,获得相应的当前帧的投影矩阵;最后使用多频带融合算法进行图像混合以改善线性加权融合算法带来的高频细节模糊。与传统算法相比,由于省去了特征提取和匹配方法,从而减少了大量的计算步骤和时间,提高了速度,增加了实用性。实验结果表明该算法具有较好的实用价值。

A video mosaic method was proposed. Firstly, phase correlation was used to roughly compute the translation offset between the first corresponding frames of two video streams, so corner match procedure was speeded up and matching stability was improved. Secondly, in the overlapped region, SIFT method was used to detect and register corners. Then, RANSAC algorithm was used to eliminate outliers to ensure effectiveness of the matched corner pairs. Singular Value Decomposition-Least Square （SVDLS） method and Levenberg-Marquardt optimization were used to robustly determine the 8 parameters transform model, For the other frames, global motion vector between consecutive frames was calculated from motion vectors included in compressed video data and projection matrix was obtained between two frames from the projection matrix of the previous frame and global motion of each input video sequence. At the last of the algorithm, a multi-band blending technique was adopted to generate the final panorama. Invalid parameters were verified by the translation offset to make Levenberg-Marquardt optimization more successful. The experiments show that the proposed algorithm exceeds existing ones in terms of matching speed and stability.