基于运动矢量分布特征的半像素快速搜索算法

Novel Fast Half-pixel Search Algorithm Based in Motion Vector Distribution Feature

针对MPEG-4视频编码中半像素全搜索方法运算量较大的缺点,提出一种基于运动矢量分布特征的快速半像素运动矢量搜索算法。试验表明最佳半像素运动矢量的分布具有一定的规律,即大多数的运动矢量分布在水平或垂直两个共扼方向上。因此在满足一定条件的前提下水平和垂直方向上的运动矢量被优先搜索,其次搜索其余的点。仿真结果表明本算法可以使半像素平均搜索点数减少71%,而编码质量和码流长度并没有明显降低或增加。

To study the conventional halfopixel full search algorithm which featured by heavy computation,a novel fast half pixel motion vector search algorithm based on the motion vector distribution characteristic is proposed to accelerate the search. Experiments have shown that the distribution of best matching motion vector observes a specific rule,i, e. , most of them locate at the horizontal or vertical direction. So the points at the horizontal or vertical direction have more priority to be the candidate than others. These points are searched firstly when some condition is satisfied, then other points may be searched. Computer simulation results have shown that the proposed fast search algorithm can reduce the search points by 71% ,while it shows only a slight degradation of image quality and bit-rate.