【作者】 黄飞跃； 徐光祐；

【Author】 Huang Feiyue, Xu Guangyou Department of Computer Science and Technology, Tsinghua University, Beijing 100084, China

【机构】 清华大学计算机科学与技术系普适计算教育部重点实验室；

【摘要】 多摄像机应用中的视频同步是一个常用而基本的步骤。大多数现有方法都依赖于特征点检测和对应,往往复杂而且难以通用。本文致力于简单视频同步方法的研究,通过对常见同步问题的描述和分析,结合基于多摄像机的动作识别的实际应用场景,提出一套简单易用的视频同步方法。该方法通过粗框架的运动特征信息来实现对应帧的匹配查找,同时采用统计方法寻求最优帧偏移量。实验证明,该方法精确度高,具有很好的实用价值。更多还原

【Abstract】 Multi cameras application is common in nowadays. In multi cameras application, camera calibration and video synchronization are two basic issues. However, video synchronization is often ignored. Even so far, only a few researches work on automatic video synchronization. Some proposed approaches depend on feature detection and point correspondence and they are complicated and hard to apply. In this paper, we aim to research on simple way for video synchronization. After a detail description and analysis on common video synchronization problem, we propose a simple method for video synchronization in a real applied scenario which is human action recognition based on multi cameras. This method uses rough motion feature information to match frames and seeks optimum frame offset with statistical method. Experiments show that results are precise enough for practical usage.更多还原