【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.更多还原

【Abstract】 Separately optimizing different video coding technologies simplifies the implementation of coding algorithms, but it results in slowdown by not taking their internal relation into consideration. An integrated algorithm is proposed to jointly improve the speed of fractional pixel interpolation(FPI) and fractional pixel motion estimation(FPME) for HEVC. Different fractional pixels are distinguished by their interpolation filters to construct a FPME algorithm with the knowledge of their interpolation cost. The FPME algorithm produces a cost effective order for refinement search which checks fractional search positions(SPs) by not only maximizing the expectation for the R-D gains but also minimizing the computational cost at the same time. On the other hand, an on-demand FPI algorithm for HEVC is also proposed to save redundant interpolation as well as reduce duplicate calculation. Experimental results show that the introduced algorithm significantly improves the overall speed of FPME and FPI with negligible coding lost.更多还原