基于二维投影变换的视频全景图拼接算法

摘要

视频序列全景图拼接技术是视频图像分析的关键技术之一.针对帧间运动模型为二维投影变换的常见视频序列,提出了一种全景图拼接算法.该算法首先采用改进的基于RANSAC的特征点匹配算法对相邻帧的运动参数进行估计,接着采用直接法对各帧与全景图间的运动参数进行精确调整,最后运用自适应中值滤波绘制出全景图.实验结果表明,该方法能够生成较高质量的全景图,基本能实现无缝连接.%Panorama mosaic is one of key techniques in video analysis. Aiming at the two-dimensional projective motion between frames used in the video sequence frequently, a mosaic algorithm for video sequence panoramic image is proposed- Firstly, the motion parameters of adjacent frame are estimated by using improved RANSAC-based matching algorithm for feature points. Secondly, the motion parameters between every frame with panorama are adjusted accurately by a direct estimation method. Finally , the video sequence panoramic image is plotted by self-adaptive median filter. Experimental results show that the proposed algorithm can generate the panoramic image with high quality.