基于多项式变换的运动估计快速算法  

Polynomial Transform Based Motion Estimation Fast Algorithm

作　　者：孙洁[1] 

机构地区：[1]深圳信息职业技术学院,深圳518034

出　　处：《计算机工程与应用》2005年第17期74-78,共5页Computer Engineering and Applications

摘　　要：主流的视频编码器普遍采用运动估计与补偿技术来提高压缩比,其中运动估计的计算复杂度高,需要占用大量的计算时间。因此,设计运动估计的快速算法对提高整个视频编码器的性能是至关重要的。此外,视频应用的实时性特点,也要求设计运动估计的快速算法。基于多项式变换的运动估计算法是论文新提出的一种块匹配运动估计算法,既保持了简单而易于硬件实现的特点,同时又极大地提高了计算效率。实验结果表明,基于多项式变换的运动估计算法的执行时间为全搜索算法的9～18%,优于其它快速算法。在噪声环境下,该算法比时间特性最好的WUS(WinnerUpdateSearch)算法以及Spiral算法快2～10倍。Motion estimation techniques play an important role in achieving high compression.It spends much time to execute motion estimation algorithm due to its high computation complexity.So,fast algorithms have to be designed to promote performance of the video code.Also,fast algorithms are important for the actual applications because of its real-time properties.A polynomial transform based motion estimation algorithm is presented with the advantages of simplicity and congruent in this paper.It promotes computation efficiency very much.Experimental results indicate that polynomial transform based motion estimation algorithm is better than what is achieved with conventional methods,its computation load is about 9～18% of full search algorithm.Experimentally,the proposed algorithm requires about 1/2～1/10 computation of WUS(Winner Update Search)and Spiral search algorithm.

关 键 词：运动估计 视频压缩 多项式变换 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]