基于运动空间相关性和时间连续性的运动估算快速算法

A Fast Motion Estimation Algorithm Based on the Spatial Correlation and the Time Continuity of the Motion Object

运动估算是视频信号的帧间预测编码中的一个重要环节 ,其效率和精度直接影响到编码器的性能 .由于全搜索算法搜索速度较低 ,而很少采用 ,故目前普遍采用三步法、交叉法等各种快速近似算法 ,但是这些算法匹配精度较低 ,而且某些情况下应用效果不好 .为解决上述算法存在的问题 ,在对视频编码中运动物体的空间相关性和时间连续性进行分析的基础上 ,给出了一种利用运动物体的空间相关性和时间连续性来进行运动估算的快速算法 .实验结果表明 ,该算法计算每个宏块运动矢量所需的平均搜索次数低于三步法 ,而匹配精度则非常接近于全搜索算法 ,并且采用该算法的编码器 。

Motion estimation is a very important part in the inter prediction technical of video coding, and the performance of a video coder largely depends on the efficiency and precision of motion estimation. The full search algorithm is seldom adopted because of its enormous complexity of computation, and now the fast algorithms of motion estimation such as three step algorithm and cross search algorithm are used widely. But the matching precision of these algorithms is low and in some cases they can not produce a good result. This paper first discusses the spatial correlation and time continuity of the motion object in video coding. Then, a new fast motion estimation algorithm based on the spatial correlation and time continuity of the motion object is presented. The average search times for computing motion vectors per macroblock of this algorithm are less than those of the three step algorithm and the matching precision of this algorithm is very close to that of the full search algorithm by experiments. Moreover, the output bits of the coder using this algorithm are less than those of the coder using the full search algorithm.