摘要
提出了一种采用多个六边形模板进行快速块匹配运动估计的新方法。该算法(MPMVFAST)充分利用了序列图像中运动矢量场的中心偏置特性和时空相关性,根据运动类型自适应的选择搜索策略和搜索起始点,结合垂直和水平方向的两个六边形搜索模板,并采用内部简易搜索算法,同时利用有效的判断中止准则结束搜索。与现有算法相比,该算法有效地减少了搜索点数,搜索精度接近于全搜索法。
In this paper,we propose a novel motion estimation algorithm,the Muhipattern-based Motion Vector Field Adaptive Search Technique（MPMVFAST）,It is composed of the prediction of initial search points,adaptive search modes among horizontal hexagon search,vertical hexagon search and small diamond search for video sequence with different motion,a fast inner search,and an effective half-stop criteria.Experiments show that MPMVFAST could obtain good PSNR performance as well as low computation cost,
出处
《计算机工程与应用》
CSCD
北大核心
2006年第8期66-69,共4页
Computer Engineering and Applications
