基于起点预测的自适应交叉 - 准菱形运动估计算法

提出一种可预测搜索起点的自适应交叉-准菱形搜索算法。根据序列图像中运动矢量的交叉-中心偏置分布特性和矢量间的时空相关性,设计了一种交叉-准菱形搜索模板,并融合搜索起点预测、半途中止准则和自适应搜索模式等技术,在保证搜索准确性的同时,大幅度提高了运动估计的速度。实验结果表明,算法对各种类型的运动序列都有很强的自适应性,其平均搜索速度是新三步法的3.56倍、四步法的3.26倍、菱形法的2.71倍,搜索准确性也优于经典的快速运动估计算法,尤其对于大运动序列,具有更明显的优势,提高了视频压缩中现有的运动估计算法的性能。

Adaptive cross-quasi-diamond search algorithm based on initial search point estimation

A predictive adaptive cross-quasi-diamond search algorithm (PACQDSA) is presented. Based on the cross-center-biased motion vector distribution characteristic of the real-world sequences and high space-time correlation of adjacent blocks' motion vectors, a cross-quasi-diamond search pattern was designed, and some effective techniques were used, which are composed of the prediction of initial search point, half-stop criteria and adaptive search modes. Experiments show that the algorithm is able to fit for all types of video sequences adaptively in spite of the degree of the motions. Compared with other algorithms, it is 3.56 times faster than NTSS, 3.26 times faster than FSS and 2.71 times faster than DS in average. And it is better than the traditional fast motion estimation algorithms in terms of PSNR, especially for the sequences with large motion. Therefore, the algorithm improves the performance of existing motion estimation algorithms.