AVS-M帧间预测模式快速选择算法的研究

Research on Fast Inter-Prediction Mode Decision for AVS-M

AVS-M是我国拥有自主知识产权的数字音视频编码系列标准的移动视频标准。AVS-M采用率失真优化(RDO)技术进行帧间模式的选择,增加编码过程中宏块进行模式选择的计算复杂度。本文提出一种快速的帧间模式选择算法。该算法首先对SKIP模式提前判决,再采用二分搜索法对帧间模式的选择,对于宏块层的模式划分采用MAD(平均绝对差值)方法,对亚宏块层的模式利用视频序列的时域、空域相关性来划分。仿真结果表明,该方法保持相同视频质量的情况下,提高了编码速度。

AVS-M is a video coding standard for mobile application,which is part of the domestic Audio Video-coding Standard（AVS）series standards.AVS-M employs rate distortion optimization（RDO） techniques in inter-prediction mode decision,but it increases the computation complexity of mode decision procedure during the coding process.The paper proposes a fast inter-prediction mode decision algorithm.Firstly,early skip mode decision is employed,Secondly,inter mode decision based on binary search methods,division of macro block level makes use of the MAD（Mean Absolute Differences） method,and according to the pace-temporal correlation,the sub macro block is divided.The experiment results show that the algorithm can reduce the encoding time while maintaining similar video quality.