一种基于运动估计的MPEG-4对象分割算法

Algorithm of object segmentation in MPEG-4 coding based on motion estimation

针对传统的视频分割算法因偏重于分割效果而导致计算量过大、实时性能差的缺点,文章提出一种在MPEG-4编码过程中通过双重筛选和利用运动估计结果提取运动对象的算法。此算法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分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,该算法不仅取得了理想的分割效果,还大幅度缩减了计算量,明显提高了实时性。

Facing the problems of large amount computation and poor real-tlme performance in traditional video segmentation algorithm, this paper puts forward a new algorithm which extracts the moving region on the basis of diplex-filtration mechanism and motion estimation in MPEG-4 coding. This algorithm implements first-layer filtration using background-priority rule in the process of motion estimation, then implements second-layer filtration adopting a self-threshold of the minimum SAD in order to obtain the good mask of the moving region. At last, using morphology filter and filling technique to eliminate noise, the accurate matrix of the motion regions are obtained. The algorithm is tested both on the hardware and software platform. The result shows that the algorithm not only has a satisfied segmentation effect, but also decreases the amount of computation and processing time.