H.264中多模式运动估计算法优化

针对H.264标准的多模式运动估计算法编码模式复杂、计算量大的不足,通过对H.264参考模型JM10中运动估计算法的分析,提出了一种利用时间、空间相关性的模式选择算法和结合一维绝对差值和(SAD,Sum of Absolute Difference)的高效匹配准则,将串行全搜索的运动估计算法改进为并行部分搜索的运动估计算法,把一维SAD和二维SAD匹配准则结合使用,从模式选择和编码速度两个方面对原算法进行优化.实验结果表明:与参考模型JM10的运动估计算法相比,提出的优化算法在恢复质量(用峰值信噪比表征)平均下降0.03 dB、码速率增加不超过1.5%的前提下,编码速率提高20%~30%,表明了该算法在恢复质量略有下降的情况下明显提高了编码速度.

Optimization of multi-mode motion estimation algorithm in H.264

The multi-mode motion estimation algorithm of H.264 standard have the problems of complicated coding mode and much computing.Based on analysis of motion estimation algorithm from H.264's reference model JM10,an original mode selecting algorithm using time or space relativity and an efficient matching rule linking with one-dimension sum of absolute difference(SAD) were designed.It optimized the former algorithm from mode selecting and coding speed by improving serial full searching motion estimation algorithm to parallel partial searching motion estimation algorithm and using one-dimension SAD matching rule combined with classical two-dimension SAD matching rule.The experiment results show that comparing with reference model JM10,the optimized algorithm makes coding rate increase 20%~30% but peak-value of signal to noise ratio(PSNR) just drop 0.03 dB and code speed just increase 1.5%.It means that this scheme increased coding speed with little quality fall of the reconstructed image.