摘要
针对当前模式选择中复杂性较高的问题,通过分析实时视频监控序列图像模块分布情况,提出了一种基于宏块运动区域特征和运动残差信息的快速模式选择算法。该算法首先将16×16块划分为四个8×8块,依据运动特征的不同将宏块划分为静态块SMB和动态块AMB;然后再利用编码块模式CBP和运动矢量残差MVD信息来作为判断准则选择较小范围的模式进行率失真代价计算。实验结果表明,对于实时视频监控序列图像,该快速模式选择算法可以在视频质量几乎没有损失的前提下,编码速度比H.264中的算法平均提高83.96%,更加适合于视频监控序列图像的编码。
In order to overcome the disadvantage of high computational complexity,with the analysis of video sequences mode decision in the real-time video surveillance system,a fast mode selection algorithm based on motion region and motion vectors difference of macroblock is proposed.Firstly a 16x16 macroblock is divided into four 8x8 blocks and it is classified into two types：SMB（Static Macroblock） and AMB（Active Macroblock） according the motion information of different regions.Then the proposed algorithm utilized a rule based on Coded Block Pattern（CBP） and Motion Vector Difference（MVD） to decide a minimum range of mode selection.The experimental results show that there is about 84.30% encoding time saved compared with the original H.264 reference software for the video surveillance sequences with almost no loss of video quality,and the proposed mode selection are very suitable for video surveillance compared.
出处
《微计算机信息》
2010年第25期1-3,共3页
Control & Automation
