摘要
立体视频对象分割是新一代视频编码、视频检索、互联网多媒体交互等新兴领域的关键技术.该文提出了一种融合多种信息的立体视频对象分割算法,首先采用基于边缘特征约束的视差估计方法,获得了可靠的视差场,再利用视差信息对视频对象进行了初始分割,在此基础上结合帧间变化检测的信息对视差分割结果进行修正.实验结果表明,本算法对有重叠的多对象具有良好的分割效果,得到了较精确的视频对象.
Stereo video object segmentation is a key technology in the new generation video coding, video browsing, and Internet multimedia interacting, etc. In this paper, a stereo video object segmentation algorithm that combines various sources of information is proposed. Estimation of disparity based on edge character restriction is ilrst adopted to obtain reliable disparity scene. segmentation. Change detection information is then integrated Experimental results show effectiveness of the algorithm in the The disparity information is used to initialize to modify boundaries of disparity segmentation. case of multiple object overlapping.
出处
《上海大学学报（自然科学版）》
CAS
CSCD
北大核心
2006年第2期116-119,124,共5页
Journal of Shanghai University:Natural Science Edition
基金
上海市科委自然基金资助项目(04ZR14056)
上海市教委发展基金资助项目(04AB56)
上海市重点学科建设项目(T0102)
关键词
立体视频
对象分割
视差
变化检测
stereo video
object segmentation
disparity
change detection
