LPPDS:基于线性预测的准菱形搜索算法

为了降低快速运动估计算法的计算复杂度，提出了一种新的块匹配运动估计算法，称为基于线性预测的准菱形搜索算法.该算法运用小菱形搜索模板，先以预测到的最小失真点作为搜索中心进行一轮小菱形搜索，然后采用线性预测确定下一轮搜索的最佳中心点，从而更快地找到整像素最优点，减少整像素搜索时间；同时该算法能很好地与目前多种分像素搜索快速算法结合使用，从而降低整个运动估计的计算时间.实验结果表明，该算法在不牺牲图像质量和压缩效率的基础上使整像素搜索时间相对于菱形搜索算法平均减少了约15%.

LPPDS: line prediction based pseudo-diamond search algorithm

In order to reduce the computational complexity of fast motion estimation,a new block-matching algorithm called line prediction based pseudo-diamond search(LPPDS) algorithm was proposed,in which the small diamond search pattern(SDSP)was adopted.The LPPDS algorithm chooses the predicted point with the minimum distortion as the initial search point to apply one step of SDSP search,and performs the line prediction to find the best center points of each step of diamond search.So that the best integer position can be found in fewer steps and the time used in integer pixel search is reduced as a result.At the same time,the new algorithm can be used with many fractional pixel search algorithms to reduce the computational load of the whole motion estimation.Simulation results demonstrated that the proposed LPPDS algorithm provides computation reduction in average of about 15% in integer pixel search compared with diamond search (DS) algorithm without sacrificing picture quality and compression ratio.