摘要
针对移动缓慢的视频对象,提出一种改进时空域联合的视频对象分割算法。时域上通过改进高阶统计算法,利用累积帧差的递归高阶统计算法检测出视频序列中视频对象运动区域。空域上利用Canny算子获得较精确的单帧目标边缘,进行了时域和空域的融合,得到精确运动对象模板后提取运动对象。实验结果表明,该算法与传统递归高阶统计算法相比能精确地对移动缓慢的视频对象进行提取,有一定的理论意义和适用性。
An improved video object segmentation algorithm based on temporal-spatial information was presented in the case of slow-moving video objects. In temporal domain, by improving higher-order statistics algorithm, the recursive higher-order statistics algorithm of accumulated frame difference was used to detect the video object moving area in video sequences. In spatial domain, Canny operator was used to obtain a more precise single frame edge. Finally, by the integration of temporal domain and spatial domain, moving objects were extracted after obtaining accurate moving objects target template. The experiments show that the algorithm can extract the slow-moving video objects more accurately than traditional recursive algo- rithm higher-order statistic algorithm.
出处
《重庆邮电大学学报（自然科学版）》
北大核心
2010年第5期660-664,共5页
Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition)
基金
国家"863"计划项目(2007AA12Z226)~~
关键词
视频对象分割
累积帧差
递归高阶统计算法
CANNY算子
video object segmentation
accumulated frame difference
recursive higher-order statistics algorithm
canny algorithm
