一种基于动态规划的视频分割方法
Dynamic Programming Based Video Segmentation
-
摘要: 提出一种基于动态规划的自动视频分割方法,该方法由运动区域识别、运动边界提取以及轮廓连接三步组成;同时提出一种基于动态规划的轮廓连接新算法,该算法能在运动区域中跟踪物体的运动边界,并同时检测出场景中位于非运动区域的静态物体边界.文中的方法在500 MHz的Pentium III PC机上实现,当系统运行在由USB摄像头采集的、图像帧大小为320×240象素的视频流上时,进行帧捕捉、视频分割和显示结果达到了15 Hz的处理速度,符合实时性的要求.Abstract: Video segmentation consists of three phases: Motion region identification,motion edge extraction and contour linkage.New algorithms for contour linkage are proposed to track the object’s motion edges as well as detect the disappearance of existing edges of the object in the scene.The proposed method is implemented on a 500MHz Pentium III PC,and the system runs on video streams with a frame size of 320×240pixels captured by a USB video camera.The process of capturing,video segmentation and result showing can reach the speed of 15Hz, which meets the requirement of real-time process.
下载: