基于视频片断的全景图拼接

Panorama Stitching in Video Sequences

视频拼接技术作为计算机视觉和计算机图形学两个学科上的交叉研究领域,在近年来得到了极大的关注,成为这两个互补领域的研究焦点之一。视频拼接技术广泛应用于虚拟现实、电影动画或特技、图像分辨率的提高,及视频压缩、视频检索等方面,有着广泛的研究价值和应用领域。利用八参数透射模型解决视频中相同场景图像之间的坐标变换问题,通过光流法求得模型的八个参数并利用求得的参数模型有效地生成全景图。实验结果显示,算法在保持较好拼接效果的同时能降低时间复杂度,适合运用在对实时性要求较高的场合。

Video mosaic has been active area of research in recent years. It becomes a major focus of attention combining two complementary fields： computer vision and computer graphics. Video mosaic has wide applications, such as virtual reality, image resolution enhancement, video compression, video indexing and so on. Using transmitting model, resolve the problem of coordinate transforrn between two images of the same scene in a video. By optical flow, figure out the 8 parameters of the model and use them to generate the panorama effectively. The experimental results show that the algorithm can reduce the rate of time complexity and maintain a relatively good stitching impression at the same time, which is fit for real - time processing.