摘要
视频对象分割是MPEG(Moving Picture Experts Group)-4视频编码的关键技术.针对常用方法的不足,提出了1种新的视频对象分割算法.形态学滤波后利用分水岭算法实现空域分割,时域分割采用二次帧差求交集的方法获取运动区域,然后时空结合.仿真结果表明该算法简单、快速、准确.
Video object segmentation is a key technology in MPEG-4 video coding. A new method is proposed, in which, a spacial segmentation is made by the watershed method after morphological filtering, and the moving regions are obtained by the intersection of three twice-frame-diffemnce images. Then, the temporal information with spacial inforrnation is intcgrated. The proposed method proves to be simple, fast and accurate.
出处
《山东大学学报（工学版）》
CAS
2005年第5期78-82,共5页
Journal of Shandong University（Engineering Science）
关键词
形态学滤波
分水岭
运动区域
二次帧差
morphological filter
watershed
moving regions
twice-frame-difference
