基于转移概率的多视点视频快速模式选择算法

针对多视点视频编码(MVC)中部分大宏块模式(SKI P/DIRECT模式,帧间16×16模式)相对于其它模式的计算复杂 度低,并且在编码所确定的最优模式中占有相当高比重的特点,本文提出一种基于转移概率 的多视点视频 快速模式选择算法。首先根据多视点视频当前编码宏块对应时间和视点间宏块及周围宏块的 位置几何关系, 建立宏块模式参考模型；其次比较参考模型中模式的时间相似度和视点间相似度；最后计算 模式之间的转 移概率,确定提前结束大宏块模式选择过程的阈值。实验结果表明,本文所提出的快速算法 平均节约MVC时间80.93%,编码质量平均下降0.04dB,码率平均增加0.27%。

Fast mode decision for multiview video coding based on transition probability

The multiview video coding (MVC) adopts the exhaustive mode decision to achieve high efficiency,at the price of extremely high computational complexi ty.Compared with the macroblock (MB) mode, the large MB mode (SKIP/DIRECT,inter 16×16) occupies most of the MB mode propor tion,while its estimation consumes much less computational complexity than that of other modes.By using t emporal and interview correlation between current MB and its adjacent MBs,a fast mode decision for MV C based on transition probability is proposed in this paper.By utilizing the geometric relationship b etween the current MB and the corresponding coded MBs in the temporal and the interview frames,the MB mode re ference model is firstly established.Compared with the temporal and interview mode similarity of the mod e reference model,the transition probability between two modes is secondly calculated.Lastly,the th reshold is obtained to terminate the process of the large MB mode decision.The experimental results show t hat compared with the exhaustive mode decision in MVC,the fast mode decision algorithm for MVC can ac hieve a reduction of the computational complexity by 80.93% in average encoding time,while incurring only 0.04dB loss in peak signal-to-noise ratio (PSNR) and 0.27% increment in th e total bit rate (BR).