一种自动分割视频对象的新方法

提出了一种视频对象自动分割及跟踪方法。该算法首先进行相邻帧差分检测和Canny边缘提取,得到相邻帧差分边缘模板;其次在相邻帧差分边缘和当前帧边缘的基础上,检测出视频对象的初始边缘模板。然后分为快变和慢变两部分进行跟踪,并对视频对象边缘模板更新以适应对象的运动;最后根据跟踪更新出的视频对象二值边缘模板,通过文中提出的填充技术及模板轮廓修正得到精确的视频对象模板提取出视频对象。实验验证表明,该算法对目标的整体运动和局部形变都有较强的适应性。此外,算法还具有简便、准确性高的特点。

A New Method of Automatic Segmentation of Video Object

A full automatic segmentation of video object and tracking scheme based on edge features is proposed in this paper.First of all,change detection and Canny edge detection are applied to get accurate difference edge mask of the inter frame.On the basis of difference edge mask and current edge mask,a binary edge model of the video object is initialized.It is then tracked and updated throughout the sequence to accommodate rotation and changes in shape of the tracked object.At last,these binary models are used to guide the actual video object plane(VOP) extraction by a two-step filling technique.The algorithm is robust to the entire motion and local deformation of object.Experiment results and object evaluation demonstrate that the proposed algorithm is computationally efficient and accurate.