首页 | 本学科首页   官方微博 | 高级检索  
     

基于假设检验和区域合并的视频对象分割
引用本文:刘志,杨杰,彭宁嵩.基于假设检验和区域合并的视频对象分割[J].数据采集与处理,2004,19(2):124-129.
作者姓名:刘志  杨杰  彭宁嵩
作者单位:上海交通大学图像处理与模式识别研究所,上海交通大学图像处理与模式识别研究所,上海交通大学图像处理与模式识别研究所 上海,200030,上海,200030,上海,200030
摘    要:提出了基于假设检验和区域合并的视频对象分割算法。初始分割采用分水岭算法,接着根据颜色相似性进行区域合并,然后利用光流场估计和全局运动估计获得全局运动的残余误差,最后以各个区域的残余误差数据进行假设检验和小区域验证来确定运动区域,通过组合所有的运动区域即可分割出具有准确边缘的完整视频对象。对MPEG-4测试序列的实验结果表明了本算法具有良好的分割性能。

关 键 词:视频编码标准  视频对象  视频序列  视频对象分割  假设检验  区域合并  MPEG-4

Video Object Segmentation Based on Hypothesis Testing and Region Merging
LIU Zhi,YANG Jie,PENG Ning-song.Video Object Segmentation Based on Hypothesis Testing and Region Merging[J].Journal of Data Acquisition & Processing,2004,19(2):124-129.
Authors:LIU Zhi  YANG Jie  PENG Ning-song
Abstract:A video object segmentation algorithm based on hypothesis testing and region merging is presented. The watershed algorithm is used to obtain an initial segmentation,and then the region merging is performed by the color similarity. The global motion residual error is calculated by using the result of optical flow and global motion estimations. Hypothesis testing is performed on those error data of each region to decide whether it is a moving region or not,and small regions need a further validation. Finally,all moving regions are grouped to obtain the whole video objects with accurate boundaries. Experimental results for the MPEG-4 test sequences demonstrate the excellent segmentation performance of the algorithm.
Keywords:video object segmentation  hypothesis testing  region merging  watershed segmentation
本文献已被 CNKI 维普 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号