【摘要】 H.264/AVC编码采用可变块尺寸运动估计,具有更好的压缩性能,但计算复杂度急剧增加,限制了其应用范围。笔者将MVFAST算法思想引入H.264/AVC,利用相邻块的相关性和模式应用概率,加快分割模式选择速度,并选定初始运动矢量,根据其运动类型选用不同的搜索模板进行块匹配,大大减少了运动估计的运算量。实验结果表明,在不影响图像质量和输出码率的前提下,该算法显著减少了运动估计时间,有效提高了编码速度。更多还原

【Abstract】 H.264/AVC has better performance of compression because of using motion estimation of variable block size,but it also enlarge computational complexity,which restricts the application range of H.264/AVC.An ameliorative algorithm named MMVFAST(Multi-Motion Vector Field Adaptive Search Technology) is proposed through the improvement on MVFAST in this paper,the selection of block pattern can be sped up by use of the coherence of contiguous blocks and the applied probability of block modes,which can also confirm original motion-vectors,and then,different search pattern is used for block-matching according to the motion type of current block.Accordingly,this algorithm can simplify the computation of motion estimation greatly.Experimental results show that this algorithm can retrench time of motion estimation and improve speed of encoding observably in the pre-requisite of the fact that the image quality and output rate keep invariant.更多还原