Low complexity algorithm for sub-pixel motion estimation of HEVC
Low complexity algorithm for sub-pixel motion estimation of HEVC
复制标题
DOI：
10.1109/icspcc.2013.6664018
复制
发表时间：
2013-11
期刊：
影响因子：
--
通讯作者：
Tsuyoshi Sotetsumoto;Tian Song;T. Shimamoto
中科院分区：
文献类型：
--
作者：
Tsuyoshi Sotetsumoto;Tian Song;T. Shimamoto
下一代视频编码标准HEVC（High Efficiency video coding）引入了许多高效的算法来提高编码效率。然而，运动估计的计算复杂度仍然是其实现的最大瓶颈。本文提出了一种分数像素运动估计（FPME）的复杂度降低算法。首先，提出了一种基于阈值的FPME快速终止算法。其次，该算法分别针对1/2像素和1/4像素的运动估计设计了几种原始搜索顺序和搜索模式。这些模式是基于每个搜索点及其周围整数搜索点的概率。仿真结果表明，该算法在几乎没有编码效率损失的情况下，成功地减少了38%的搜索点。
The next generation video coding standard, namely HEVC(High Efficiency Video Coding) introduced many efficient algorithms to enhance the coding efficiency. However, the computational complexity of motion estimation is still the most bottle neck for its implementation. In this paper, we propose a complexity reduction algorithm for fractional-pixel motion estimation (FPME). Firstly, a fast termination algorithm is proposed to fast terminate FPME using a threshold. Next, the proposed algorithm designs several original search orders and search patterns for 1/2 pixel and 1/4 pixel motion estimation, respectively. These proposed patterns are on the basis of the probability of each search point and its surrounding integer search points. Simulation result shows that the proposed algorithm succeed in reduction of search point for up to 38% with almost no coding efficiency loss.