论文部分内容阅读
视频分割是视频结构化和检索的重要技术 ,目前主要通过镜头分割得到。但许多应用需要对镜头根据摄像机运动作分割 ,现有技术不能满足应用实时需要。提出了一种对P帧运动向量稳健分析计算帧间运动并进行分割的方法。与传统光流法、M估计法比较 ,该算法只在压缩域上进行 ,不需迭代 ,具较强实时性。文中阐述了方法的基本思想、理论依据与实现 ,实验表明 :该方法具有很好的效果
Video segmentation is an important technology for video structure and retrieval, which is mainly obtained through the lens segmentation. However, many applications require that the lens be segmented according to camera motion, and the prior art does not meet the real-time needs of the application. A method of robustly analyzing P-frame motion vectors and segmenting them is proposed. Compared with the traditional optical flow method and M estimation method, the algorithm is only carried out on the compressed domain, without iteration, with strong real-time. The article expounds the basic idea, theoretical basis and realization of the method, the experiment shows that: the method has very good effect