基于多帧灰度差异的视频对象分割  

Segmentation of Video Object Based on Change of Multi-frames' Gray Value

作　　者：陈国锋[1] 何南忠[1] 施保昌[1] 

机构地区：[1]华中科技大学数学系,武汉430074

出　　处：《计算机工程与应用》2006年第12期66-67,88,共3页Computer Engineering and Applications

基　　金：国家自然科学基金资助项目(编号:60473015)

摘　　要：在新一代MPEG-4视频编码标准中,为了支持面向视频对象的编码和实现基于内容的应用,自动、快速地从视频场景中提取视频对象已成为一项关键技术,为此文章提出了一种基于多帧灰度差异的视频对象分割新算法。该算法首先采用多帧序列图像灰度差异的4次高阶统计量假设检验(HigherOrderStatistics,HOS),自动提取运动对象模板;然后利用数学形态学对模板作平滑和内部填充处理;最后将模板投影到原始灰度图,得到运动对象。实验结果表明,该方法能比较有效地分割视频对象。In the new MPEG-4 coding standard,a fast and automatic segmentation of video object plays a key role in supporting object-oriented coding and enabling content-based functionalities.In this paper,a new algorithm for extracting moving object is proposed.First,a binary mask is drawn by fourth-order statistics detection method based on the gray differences between multiply frames.Then,we use morphologic methods to smooth and fill the mask.Finally,the moving object is extracted by projecting the mask on the original gray picture.Experimenttal results demonstrate that this method is effective.

关 键 词：运动对象 视频对象分割 高阶统计量 形态学 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]