H.264中UMHexagonS运动估计算法的改进

发布时间：2018-03-11 22:09

本文选题：H.　切入点：运动估计　出处：《计算机应用》2014年06期 　论文类型：期刊论文

【摘要】：针对H.264中UMHexagonS算法进行研究,提出了改进的快速运动估计算法。首先对UMHexagonS算法中的固定搜索窗口、非对称十字搜索、5×5的小矩形螺旋搜索、非均匀多层次六边形格点搜索和扩展的六边形搜索进行研究;然后提出了相应的减少搜索点数的优化算法,这些优化算法分别称作动态搜索窗口、自适应十字模型搜索、方向性的3×3小矩形搜索、基于预测方向的密集搜索和改进的扩展六边形搜索,这些优化算法形成了自适应模型方向搜索(APDS)算法。对各种类型的测试序列进行实验,与UMHexagonS算法相比,结果表明APDS算法在重建视频质量和码率基本相当的条件下,运动估计时间平均减少了29.64%,每形成一个运动矢量平均节省搜索了21.64个点数;而且运动估计效率的提升,最终增强了编码器的实时性能。
[Abstract]:Based on the research of UMHexagonS algorithm in H.264, an improved fast motion estimation algorithm is proposed. Firstly, the fixed search window in UMHexagonS algorithm, the asymmetric cross search for 5 脳 5 small rectangular helical search, is used to search. The non-uniform multi-level hexagonal lattice search and the extended hexagonal search are studied, and the corresponding optimization algorithms to reduce the number of search points are proposed. These optimization algorithms are called dynamic search window, adaptive cross model search, respectively. The directional 3 脳 3 small rectangle search is based on the intensive search of prediction direction and the improved extended hexagonal search. These optimization algorithms form an adaptive model direction search (APDS) algorithm. Compared with the UMHexagonS algorithm, the results show that the APDS algorithm reduces the average motion estimation time by 29.64 points and saves 21.64 points per motion vector under the condition that the reconstructed video quality and bit rate are basically the same. Moreover, the efficiency of motion estimation improves the real-time performance of encoder.
【作者单位】：上海海事大学信息工程学院;
【基金】：国家自然科学基金资助项目(61171126)
【分类号】：TN919.81

【相似文献】