在视频编码系统中,半像素精度的运动估计虽可以明显地改善编码效果,但也因此增加了不少运算量。为了降低运算量和提高半像素运动估计的速度,提出了一种全新的半像素搜索算法(half-pixel motion estimation based on linear prediction,简称BLPHME),其关键思想是通过分析整像素搜索和半像素搜索结果之间的相关性来建立一个线性模型,通过动态调整判决门限,以便预测并跳过那些不能从半像素块匹配搜索中得到好处的块。实验结果表明,该算法不仅可以明显地降低运动估计的运算量,同时还能得到与传统算法非常接近的图像质量和码率。此外,该算法还可以和基于整像素和半像素的快速运动估计算法一起使用,以进一步降低运算量。

In video coding system,motion estimation at half-pixel accuracy can obviously enhanced coding efficiency compared to the motion estimation at integer-pixel accuracy only.However it requires more computation at the same time.In order to reduce the computation and while enhance the speed of motion estimation at half-pixel,we proposed a novel half-pixel motion estimation algorithm named half-pixel motion estimation based on linear prediction(BLPHME) for video coding in this paper.The key point of the algorithm is setting up a linear model by analyzing the relativity between the results of integer-pixel search and half-pixel search.Then we can modify the threshold dynamically for each frame.By doing so,it can skip over the blocks that can't be benefited from the half-pixel search based on the block size.Experimental results show that significant reduction in computation of motion estimation is achieved together with the increment in bit rate by using the proposed method,but without obvious increment in bit rate and visible loss in video fidelity and increment in bit rate.Moreover,the proposed algorithm can be combined with any of the fast motion estimation algorithm at integer-pixel or half-pixel accuracy to further reduce the complexity.