HEVC的高效分像素运动补偿

发布时间：2018-07-05 09:36

本文选题：HEVC + 分像素运动估计　；参考：《软件学报》2017年08期

【摘要】：现有的视频编码方法单独针对不同的编码技术进行优化,虽然简化了算法的实现,但却忽略了这些技术之间的内在关系而局限了性能的进一步提升.尝试针对HEVC中的分像素插值和分像素运动估计,建立一种整合的优化算法,同时提升两者的计算速度.首先,通过细分不同分像素的插值方法,把每个分像素的插值代价融合在分像素运动估计的搜索中,并构建一种性价比优先的分像素运动估计方法.该运动估计方法的每一步搜索都按照最小化计算代价和最大化率失真收益的原则进行,保证以最小的计算消耗获得最优的性能.其次,为了配合该方法,针对HEVC的特性构建了一种分区域分像素集的分像素插值算法.该插值算法在分像素运动估计过程中仅动态地计算所用到的分像素,解决了现有插值算法无法同时降低重复计算和冗余计算的问题.实验结果表明,该整合算法可以在几乎不产生率失真性能损失的情况下,大幅度地加快分像素运动估计和插值的速度.
[Abstract]:The existing video coding methods are optimized for different coding techniques, which simplify the implementation of the algorithms, but ignore the inherent relationship between these techniques, and limit the further improvement of performance. This paper attempts to establish an integrated optimization algorithm for sub-pixel interpolation and sub-pixel motion estimation in HEVC. Firstly, by subdividing different sub-pixel interpolation methods, the interpolation cost of each sub-pixel is fused into the sub-pixel motion estimation search, and a performance-to-price priority sub-pixel motion estimation method is constructed. Every step search of the motion estimation method is carried out according to the principle of minimizing the computational cost and maximizing the rate-distortion benefit, which ensures the optimal performance with the minimum computational cost. Secondly, according to the characteristics of HEVC, a sub-pixel interpolation algorithm for sub-region pixel set is proposed. This interpolation algorithm only calculates the sub-pixel dynamically in the process of sub-pixel motion estimation, which solves the problem that the existing interpolation algorithm can not reduce the redundant computation and duplicate computation at the same time. Experimental results show that the proposed algorithm can greatly accelerate the speed of sub-pixel motion estimation and interpolation without rate distortion performance loss.
【作者单位】：广东金融学院计算机科学与技术系;中山大学软件学院;
【基金】：国家自然科学基金(61173081) 广东省自然科学基金(2014A030313662)~~
【分类号】：TN919.81

【相似文献】