论文部分内容阅读
本文提出了一种运动估计的快速预测搜索算法(PSA)。该算法首先用当前块的三个邻近块运动矢量的线性加权来得到预测矢量,然后以预测点为起始点,采用3×3的搜索窗进行搜索步长为1的移动窗搜索,直到搜索到达搜索域的边界或搜索的局部最小点位于搜索窗的中心时停止。该算法由于利用了序列图象的实际运动矢量与预测矢量之间距离的空间分布特性一中心偏置分布特性和时间上的相关特性,并采用了中止判决准则,可以明显地减少搜索次数。仿真表明这种算法减少了搜索范围和搜索次数,提高了搜索效率,降低了运动估计的计算复杂性。本文还详细地给出了PSA算法与其它常用快速算法的比较结果。
In this paper, a fast predictive search algorithm (PSA) for motion estimation is proposed. The algorithm first uses the linear weighting of the motion vectors of the three neighboring blocks of the current block to obtain a prediction vector, and then uses a 3 × 3 search window to search for a moving window with a search step of 1 using the prediction point as a starting point until the search reaches The boundary of the search field or the searched local minimum point stops at the center of the search window. The algorithm can effectively reduce the number of search due to the use of a center-bias distribution and temporal correlation characteristics of the spatial distribution of the distance between the actual motion vector and the prediction vector of the sequence image and the use of the stop criterion. Simulation shows that this algorithm reduces the search range and search times, improves the search efficiency and reduces the computational complexity of motion estimation. This article also gives a detailed comparison between the PSA algorithm and other commonly used fast algorithms.