摘要
文章提出了一种新的在MPEG-4编码过程中提高实时性的算法。算法采用双重筛选机制,充分利用了运动估计的结果提取运动对象。首先在运动估计过程中加入背景优先准则进行第一重筛选;然后根据运动估计后宏块的最小平均绝对差值利用区域自适应门限进行第二重筛选,从而获得运动区域掩模;最后对掩模进行形态滤波和填充消除噪声点,最终得到运动区域。算法分别在硬件和软件平台上进行测试,结果表明提出的算法得到了理想的分割效果,较传统的在编码前进行运动对象提取的方法实时性大大提高。
A new algorithm is proposed, which improves the real-time capability in the process of MPFG-4 coding. The algorithm uses diplex-filtration mechanism, and takes full advantage of the result of motion estimation. First, some relevant rules are added to the process of motion estimation for the first filtration. Then the self-adapting filtration is carried through according to the minimum Sum of Absolute Difference (SAD), and the mask of the moving region is obtained. Finally, after the morphology filtering and noise removing, the accurate matrix of the motion regions is obtained, and the moving video object is thereby gained. The experimental results show this algorithm gets perfect segmentation, and improves the real-time capability greatly when compared with the traditional method of extracting the moving video object before coding.
出处
《中兴通讯技术》
2007年第2期42-46,共5页
ZTE Technology Journal
关键词
运动对象
运动估计
自适应门限
形态滤波
moving video object
motion estimation
self-adapting threshold
morphology filter