中文摘要：

提出了一种基于宏块级运动预检测的模式选择算法.采用低复杂度的联合运动检测准则对当前宏块运动程度进行评估.根据当前宏块与空间相邻宏块的运动程度将宏块分级,并采用不同的编码模式判决方法.该算法能较好地区分背景噪声与运动物体,并尽可能保留运动宏块细节.对一些典型监控场景视频序列的仿真实验结果显示,该算法平均节约了75.2%的编码时间.与H.264参考软件中的模式选择算法相比,该算法不但节约了1.31%的平均码率,而且平均峰值信噪比提高了约0.08dB.

英文摘要：

On the basis of early macroblock level motion detection,a novel mode decision algorithm is proposed,in which a low-complexity joint motion detection rule is used to evaluate the current marcoblak （MB） motion level.According to the current and spatially adjacent MB motion levels,different mode decision strategies were chosen for current MB coding.The proposed algorithm can distinguish between the background noise and the moving object.Simulations show that this approach can result in a time savings of over 75.2% for several typical surveillance sequences.Compared with the mode decision algorithm of H.264,this algorithm can reduces Bjontegaard delta bit rate by about 1.31% and increase Bjontegaard delta peak signal-to-noise ratio by about 0.08 dB on an average.