【摘要】 提出了一种新的利用MPEG-4编码过程中宏块运动估计的结果来提取运动对象的算法。该算法在运动估计过程中加入相应的准则,充分利用运动估计中宏块的运动矢量和最小平均绝对差值来获取运动区域的掩模矩阵,最后经过消除噪声点和空洞获得精确的掩模矩阵从而获得运动对象,该方法能在编码过程中进行。实验证明:利用该方法得到了理想的分割效果,较传统基于像素的运动对象提取的方法大大节省了处理时间。更多还原

【Abstract】 A new algorithm for segmentation of moving video object is proposed, which utilizes the information of motion estimation. Firstly, some rules are added in the process of motion estimation, and then the moving regions mask is obtained by macroblock mo- tion vectors and the value of SAD. At last, the accurate mask can be obtained after removing noise; therefore the moving video ob- ject can be got. The algorithm can run in the process of coding. Experimental results show that the algorithm can get satisfactory re- sult. Comparing to methods in pixel domain, it save a lot of time.更多还原